Академический Документы
Профессиональный Документы
Культура Документы
uk/wrap
A Thesis Submitted for the Degree of PhD at the University of Warwick
http://go.warwick.ac.uk/wrap/3837
This thesis is made available online and is protected by original copyright.
Please scroll down to view the document itself.
Please refer to the repository record for this item for information to help you to
cite it. Our policy information is available from the repository home page.
School of Engineering
University of Warwick
April 2010
Table of Contents
TABLE OF CONTENTS....i
LIST OF ABBREVIATIONS.v
LIST OF MATHEMATIC AND GREEK SYMBOLS.vii
LIST OF FIGURES...x
LIST OF TABLES...xiii
ACKNOWLEDGEMENTSxiv
DECLARATION..xv
LIST OF PUBLICATIONS.xvi
ABSTRACT...xvii
CHAPTER 1: INTRODUCTION..1
1.1
Overview..1
1.2
Optical Wireless Communication .....4
1.2.1 System Structure ...6
1.2.2 Optoelectronic components .9
1.2.2.1 Transmitter Optical Component...9
1.2.2.2 Receiver Optical Component.10
1.3
Project Motivation...10
1.4
Thesis Structure...12
2.4
2.5
Original Contributions42
Summary and Conclusions. ...43
5.5
5.4.1
System Structure...116
5.4.2
Adaptive Model Identification...117
5.4.3
Singleton Data Set.....117
5.4.4
2-D Recursive Data Set..118
5.4.5
Training the ANFIS Model119
5.4.6
Results Comparison..120
Summary and Conclusions...124
APPENDIX
APPENDIX II-1
APPENDIX II-2
APPENDIX III-1
APPENDIX IV-1
APPENDIX IV-2
APPENDIX IV-3
APPENDIX IV-4
APPENDIX IV-5
APPENDIX IV-6
APPENDIX IV-7
APPENDIX IV-8
APPENDIX IV-9
APPENDIX V-1
APPENDIX V-2
APPENDIX V-3
APPENDIX V-4
APPENDIX V-5
APPENDIX VI-1
APPENDIX VI-2
APPENDIX VI-3
APPENDIX VI-4
REFERENCES:..221
iv
LIST OF ABBREVIATIONS
AIr
AP
APD
AWGN
BA
BER
CDMA
CRC
DFIR
DH-PIM
DR
DPPM
DAPPM
FC
FCC
FDMA
FEC
FL
FLS
EM
FOV
FR
FS
HP
HPF
i.i.d
IM/DD
IR
IrDA
ISI
ISO
LAN
LED
LO
LOS
MAC
MAP
MDPIM
MF
MPPM
MPAPM
NRZ-OOK
advanced infrared
access point
avalanche photodiode
additive white Gaussian noise
bandwidth allocator
bit error rate
code division multiple access
cyclic redundancy check
diffuse infrared
dual header pulse interval modulation
data rate
differential pulse position modulation
differential amplitude pulse position modulation
fuzzy control
federal communications commission
frequency division multiple access
forward error correction
fuzzy logic
fuzzy logic system
electromagnetic
field-of-view
fuzzy rule
fuzzy set
hewlett-packard
high-pass filtering
independent and identically distributed
intensity modulation with direct detection
infrared
infrared data association
inter symbol interference
international organization for standardization
local area network
lighting emitting diode
logic operation
line-of-sight
medium access control
maximum a posteriory
Multilevel digital pulse interval modulation
membership function
multiple pulse position modulation
multiple pulse amplitude and position modulation
none return to zero OOK
v
OOK
OW
PAPM
PIN
PPM
QoS
RC
RCPC
RF
RS
RZ-OOK
SNR
SYNC
TDMA
TH
ToS
TSK
VoIP
XOR
on-off keying
optical wirless
pulse amplitude and position modulation
the diode with an intrinsic layer between the P and N-type regions
pulse position modulation
quality of service
repetition codes
rate-compatible punctured convolution codes
radio frequency
reed-solomon
return to zero OOK
signal to noise ratio
synchronization
time division multiple access
threshold detection
type of service
Takagi-Sugeno-Kang model
voice over IP
exclusive or
vi
<
>
!
~
^
{,}
{|}
equal to
not equal to
less than
greater than
much less than
much greater than
less than or equal to
greater than or equal to
is proportional to
absolute value
factorial (e.g. n! means the product 1 2 ... n)
probability distribution (e.g. X ~ D, means the random variable X has
the probability distribution D).
approximately equal
because
therefore
exponentiation
set brackets
set builder notation
an element of
subset
superset
if and only if
convolution
summation
product
integral
mean value of
infinity
1 , 2
A
second amplitude parameters of low frequency components
ceiling height
()
channel impulse response.
1
data rate
{}
{}
()
emitting angles
detector threshold
angle between receiver and ( )
incident angles
phase parameters of high frequency components
Dirac delta function
Gaussian noise variance
surface reflection coefficient
reflection coefficient of surface
reflector element treated as emitter
first phase parameters of low frequency components
second phase parameters of low frequency components
fraction factor between RZ-OOK and NRZ-OOK
ratio between peak and average intensity
pulse power after channel inference
membership functions
ratio of a circle's circumference to its diameter
ix
LIST OF FIGURES
CHAPTER 1
Figure 1.1 Transmission and reception in an infrared link with IM/DD..7
Figure 1.2 Classification of optical wireless link........9
CHAPTER 2
Figure 2.1 Optical wireless system diagram.........15
Figure 2.2 Equivalent channel model...........16
Figure 2.3 Capacity bounds and mutual information for continuous one-sided
exponential, Gaussian and discrete uniform PAM.18
Figure 2.4 PPM capacity on the AWGN channel, determined by Monte Carlo
simulation....19
Figure 2.5 Geometry of optical source and detector..22
Figure 2.6 Single reflection propagation model...25
Figure 2.7 Multiple reflection model....25
Figure 2.8 Propagation model distorted by multipath effects...27
Figure 2.9 Propagation model employing multipath effects.....27
Figure 2.10 Impulse response of room A (K=1, 2, 3) (Unblocked) ...29
Figure 2.11 Impulse response of room A (K=1, 2, 3) (Blocked) ...29
Figure 2.12 Ceiling bounce model......31
Figure 2.13 Background radiations with Si-photodiode responsivity.33
Figure 2.14 Typical artificial light interference time waveform and spectrum of
(a) Incandescent lamp (b) Fluorescent lamps driven by conventional
ballast and (c)Fluorescent lamp driven by electronic ballast (energy
saving lamp)....34
Figure 2.15 Sample interference waveform of incandescent lamp driven by
electronic ballast with =1A/W and =1W.37
Figure 2.16 Modulation performance in channel limited by (a) shot noise only
(b) incandescent light without HPF (c) incandescent light with
HPF..38
Figure 2.17 Modulation performances in channel limited by (a) fluorescent light
driven by conventional ballast with and without HPF (b) fluorescent
light driven by electronic ballast with and without HPF.39
CHAPTER 3
Figure 3.1 Family tree of pulse modulation schemes for optical wireless
systems.46
Figure 3.2 Comparison of (a) NRZ-OOK pulse (b) RZ-OOK pulse with duty
cycle = 0.5...47
Figure 3.3 The continuous portion of the power spectral density of OOK
scheme.48
x
Figure 3.4
Figure 3.5
Figure 3.6
Figure 3.7
Figure 3.8
CHAPTER 4
Figure 4.1 Channel impulse response (H=10m)...65
Figure 4.2 Channel impulse response according to H.....66
Figure 4.3 Normalised power and bandwidth requirement of L-PAM....71
Figure 4.4 Optimum adaptive ratio value search (L-PAM).........76
Figure 4.5 Normalised power and bandwidth requirement of L-PPM....78
Figure 4.6 Optimum adaptive ratio value search (L-PPM) ..82
Figure 4.7 Normalised power and bandwidth requirement of M-n-PAPM..84
Figure 4.8 Optimum adaptive ratio value search (M-n-PAPM)90
Figure 4.9 OOK and L-PAM SNR vs BER comparison (with L=2, 3, 4, 5)93
Figure 4.10 BER to ceiling height for OOK and 2-PAM..............94
Figure 4.11 BER to data rate for OOK and 2-PAM....95
Figure 4.12 BER to data rate for 2-PPM.97
Figure 4.13 Zoomed version of BER to data rate for 2-PPM....98
CHAPTER 5
Figure 5.1 General categories of AI....101
Figure 5.2 Block diagram of FL controlled adaptive modulation system..103
Figure 5.3 Structure of fuzzy system..103
Figure 5.4 Fuzzy logic system block diagram....106
Figure 5.5 BER variations to fuzzy set mapping....109
Figure 5.6 Fuzzy set to required level changes mapping...110
Figure 5.7 Block diagram of adaptive PAPM fuzzy system (System A)...111
Figure 5.8 Fuzzy system inputs/outputs for system A....111
Figure 5.9 Fuzzy inference process for system B...113
Figure 5.10 Block diagram of adaptive PAPM fuzzy system (System B)..114
Figure 5.11 Fuzzy system inputs/outputs for system B...115
Figure 5.12 ANFIS rule operation example..116
Figure 5.13 Comparison of single ton and 2-D recursive data set generatio..118
Figure 5.14 Singleton (a) BER variation (b) Rate value (c) Output levels and
recursive (d) BER variation (e) Rate value (f) Output levels119
Figure 5.15 ANFIS trained by BPGD on singleton data set...121
Figure 5.16 Training error of BPGD on singleton data set..121
Figure 5.17 ANFIS Trained by hybrid on singleton data set...122
Figure 5.18 Training error of hybrid on singleton data set..122
Figure 5.19 ANFIS trained by hybrid on recursive data set....123
xi
CHAPTER 6
Figure 6.1 BER and data rate performance for M-n-PPM (M=1, n=4)
modulation scheme with variable H and no ambient light
interference128
Figure 6.2 BER and data rate performance for M-n-PPM (M=1, n=4)
modulation scheme with variable ASR and c onstant ISI
(H=1m)........................................................131
Figure 6.3 BER and data rate performance for candidate adaptive M-n-PPM
modulation scheme with ASR=50 and H=1m...136
Figure 6.4 SNR to BER performance for candidate adaptive M-n-PPM
modulation scheme with ASR=50 and H=1m...137
Figure 6.5 Fuzzy system inputs/outputs for system C...138
Figure 6.6 ANFIS trained using hybrid with recursive data set (System D)..139
Figure 6.7 Training errors of system D...139
xii
LIST OF TABLES
CHAPTER 1
Table 1.1 Comparison of ISM, LMDS and FSO systems..3
Table 1.2 Comparison between radio and IM/DD infrared systems for indoor
Wireless communications......5
Table 1.3 Comparison of LEDs and LDs.10
CHAPTER 2
Table 2.1 Types of radiation and their likely effects on the human eye20
Table 2.2 Laser safety classifications for a point-source emitter.20
Table 2.3 Typical values for phase parameter and 36
Table 2.4 Typical value for amplitude and phase parameters of high frequency
components.36
CHAPTER 4
Table 4.1 Data rate degradation of OOK..70
Table 4.2 L-PAM value matrix of adaptive factors..74
Table 4.3 Comparison of adaptive and interference ratio for L-PAM..75
Table 4.4 Data rate recovery of L-PAM...77
Table 4.5 L-PPM value matrix of adaptive factors...80
Table 4.6 Comparison of adaptive and interference ratio for L-PPM..81
Table 4.7 Data rate recovery of L-PPM...83
Table 4.8 M-n-PAPM value matrix of adaptive factors...87
Table 4.9 Table 4.9 Comparison of adaptive and interference ratio for
M-n-PAM89
Table 4.10 Data rate recovery of M-n-PAM , {2,3,4}91
CHAPTER 5
Table 5.1 Modulation parameter change rate.105
Table 5.2 BER degradation mapping.....107
Table 5.3 ANFIS system training parameters....120
CHAPTER 6
Table 6.1 System parameters for adaptive M-n-PAPM modulation with
variable H and no ambient light noise ...129
Table 6.2 System parameters for adaptive M-n-PAPM modulation with
variable ASR and constant ISI (H=1m).........133
Table 6.3 Initial system parameters for adaptive M-n-PAPM (M=1, n=4)
modulation with H=1 and ASR=50............135
Table 6.4 System parameters for adaptive M-n-PAPM (M=1, n=4) modulation
with H=1 and ASR=50 using exhaustive search....135
xiii
ACKNOWLEDGEMENTS
I would like to thank my supervisors, Prof. Roger Green and Dr. Mark Leeson
for their support and encouragement over the years. I wouldnt imagine what I
can achieve without Prof. Roger Green and Dr. Mark Leeson. I would also like to
thank Prof. Fary Ghassemlooy and Dr. Christos Mias to act as my external
examiner and internal examiner. Thanks for the professional guidance you have
provided during my examinations.
Thanks goes to the following people from the School of Engineering, University
of Warwick for their assistance during my PhD: Dr. Zur Abu Bakar, Dr. Loh
Tianhong, Dr. Roberto Ramirez-Iniguez, Dr. Xiaoming Jian, Dr. Lei Sun, Dr. Lei
Xue, Dr. Philip Shepherd, Dr. Dean Hamilton and Mr. Shaobo Sun, Dr. Matthew
Higgins, Miss Harita Joshi, Mr.Bo Zhao, Ms. Yanling Zhai for being my group
mates and all the fruitful discussions.
Special thanks go to all my friends, who have supported me thought the easy and
hard times, especially Jackie Cai. Last, I would like to thank for my parents, for
their understanding, patience and continues support during my years at Warwick,
who make it possible for me to complete this PhD.
xiv
DECLARATION
This thesis is presented in accordance with the regulations for the degree of
doctor of philosophy. All work reported has been carried out by the author unless
otherwise stated. This thesis has not been submitted for a degree at another
university.
xv
LIST OF PUBLICATIONS
Journals:
1. H.F.Rashvand, Y.Zeng, R.J.Green and M.S.Leeson, "Lookup Table Error
Correcting Multiple Pulse PPM Codes for Wireless Optical Communication
Channels" IET Communications Special Issue on Optical Wireless
Communication Systems, Volume 2, Issue 1, pages 27-34, January 2008
2. Y.Zeng, R.J.Green, S.B.Sun and M.S.Leeson, "Tunable Pulse Amplitude and
Position Modulation Technique for Reliable Optical Wireless Communication
Channels" Journal of Communications, Academy Publishers, Vol. 2, No. 2, pages
22-28, March 2007
Conference:
1. Y.Zeng, R.J.Green and M.S.Leeson, "Multiple pulse amplitude and position
modulation for the optical wireless channel" the IEEE ICTON 2008 10th
International Conference on Transparent Optical Networks, Volume 4, pages
193-196 (We.C4.4), Athens, Greece, June 22-26, 2008
2.. Y.Zeng, R.J.Green and M.S.Leeson, "Adaptive Pulse Amplitude and Position
Modulation for Optical Wireless Channels" at the The 2nd IEE International
Conference on Access Technologies, pages 13-16, Abington Hall, Cambridge,
UK, 21st to 22nd June 2006,
3. Y.Zeng, R.J.Green, "Modulation Adaptive System for Wireless Infrared
Channels" at the 5th annual Postgraduate Symposium on the Convergence of
Telecommunications, Networking and Broadcasting (PGNET 2004), pages 74-77,
University of Liverpool, Liverpool, UK, 28-29 June 2004
xvi
Abstract
High-speed wireless optical communication links have become more popular for
personal mobile applications. This is a consequence of the increasing demand
from the personal information service boom. Compared to the radio frequency
domain, optical wireless communication offers much higher speeds and bit rates
per unit power consumption. As stated by the official infrared standard IrDA
optical communication enjoys much lower power consumption than Bluetooth,
with an inherent security feature while in Line of Sight (LOS) applications. There
are also drawbacks such as the infrared radiation cannot penetrate walls as radio
frequencies do and interference from the background contribute to the channel
dispersions.
Focus on the modulation aspects of the optical wireless communication, this
thesis try to improve the channel immunity by utilising optimised modulation to
the channel. Modulation schemes such as on off keying (OOK), pulse amplitude
modulation (PAM) and pulse position modulation (PPM) and pulse position and
amplitude modulation PAPM schemes have been validated. The combined power
and bandwidth requirements suggest that the adaptive modulation schemes can
provide reliability when deployed in a real time channel, resulting in improved
system performance.
As a result, an adaptive modulation technique is proposed. Extensive simulations
of severe noise distraction have been carried out to validate the new scheme. The
simulation results indicate that the new scheme can provide increased immunity
against channel noise fluctuation at a relatively low complexity. The scheme
obtained formed a basis to support reliable mobile optical wireless
communication applications.
The adaptive scheme also takes the real time channel conditions into account,
which is different from existing schemes. Guaranteed system performance can be
secured without compromising power and bandwidth efficiency. This is also a
new approach to realise reliable optical wireless links. Fuzzy logic control
module has been developed to match the adaptive pattern.
xvii
Chapter 1
Introduction
1.1
Overview
1.2
System Structure
1.2.2
Optoelectronic Components
1.2.2.1
1.2.2.2
1.3
Project Motivation
1.4
Thesis Structure
1.1 Overview
The increasing demand for bandwidth had driven researchers to explore new
technologies to accommodate more data throughput over the decades [1-7]. As the
conventional radio frequency (RF) domain becomes heavily congested, the search
for an alternative information transmission medium took priority [8-10]. Optical
wireless communication attracted considerable attention from the academic
community [10-14]. Starting from short distances and low speed experimental
links, the optical wireless communication domain became a viable addition to
communication systems, and showed promising prospects [15-21]. Suggested by
In optical communications, there were two mainstream areas: fixed optical fibre
and free space optical (FSO) links. The former found most applications in long
distance communications. For example, the optical fibres with attenuation less
than 20dB/km were demonstrated in 1970 [23]. Optical fibre gradually replaced
copper wire in consumer markets; service providers, such as the Internet service
providers (ISPs), cable television (CATV), and telephone companies already
utilised it widely [24]. To deliver the required connectivity, these service
providers faced challenges in reaching the individual customers, namely the last
mile problem [25].
The free space optical wireless link mainly been applied in short range (less than 2
kilometres) and inter-building data connections complementary to existing RF
networks. Although challenged by several competitive RF bands, including the
industrial, scientific and medical (ISM) radio bands, and the local multipoint
distribution service (LMDS) bands [29], optical wireless showed the promising
ISM Band
LMDS
Optical Wireless
Frequency
Licensed
Multipoint Topology
Cell Radius
2.4GHz
No
Omni or Sectored
8-15km
3-8Mbps per sector
(per frequency)
3Mbps peak
per user
No
No
Heavy Rain
24-40GHz
Yes
Omni or Sectored
2-3km
155Mbps per
sector
3-10Mbps per
user
No
No
Rain
30-60THz
No
Virtual Multipoint
1-2km
1.5Gbps
per user
1.5Gbps per
user
Yes
Yes
Thick Fog, Snow
High
High
Low
Medium
Medium
Medium
Downstream Bandwidth
Upstream Bandwidth
Symmetric
Protocol Independence
Fade Mechanism
Initial Investment for few
subscribers
Investment for 50-100
subscribers per cell
From the above table, the optical wireless (OW) channel surpassed the RF system
in following aspects: the downstream bandwidth per user/sector/frequency of OW
system was nearly 10 times that of the LDMS system and up to 500 times that of
the ISM system. The upstream bandwidth was similar to that of the downstream
bandwidth. In the cell radius comparison, the OW system provided the shortest
distance coverage, where ISM and LDMS systems can achieve a range which as
7.5 and 1.5 times further than the OW system respectively. Noticeably, weather
conditions had an impact on the reliability of the channel, which could affect the
transmission data rate.
rate of a few gigabits per second and higher [32]. More details of the channel type
were discussed in Chapter 2.
Table 1.2 Comparison between radio and IM/DD infrared systems for indoor
wireless communications (table adapted from [15])
Property of Medium
Bandwidth Regulated?
Radio
Yes
IM/DD Infrared
No
Yes
No
Multipath Fading?
Multipath Distortion?
Path Loss
Dominant Noise
Input X(t) Represents
SNR Proportional to
Yes
Yes
High
Other Users
Amplitude
No
Yes
High
Background Light
Power
Average Power
Proportional to
X(t) 2 dt
X(t) 2 dt
X(t) 2 dt
X(t)dt
Implication for IR
Approval not required.
Worldwide compatibility
Less coverage.
More easily secured.
Independent links in
different rooms.
Simple link design.
Limited range.
Difficult to operate outdoors
High transmitter power
requirement.
Choose waveform X(t) with
high peak-to-average ratio.
Apart from points listed in Table 1.2, another benefit to use IR over RF was from
the health concerns. Side effects caused by exposure to electromagnetic (EM)
radiation were still ongoing research topics [33]. Since human nervous system
receive and interpret information via electrical signals [34], possible carcinogenic,
reproductive and neurological effects may indeed develop due to exposure to
intense EM radiations [35].
Since the 90s, extensive research efforts had been focused on improving the
channel performance. This included modulation [36-43], coding and equalization
[44-47], diversity detection [48-50], multiple access [51], channel characterization
and modelling [52, 53], optical component design [54-56], prototype
communication links [16, 57, 58] etc. There were activities also welcomed by the
beginning of the official interest group, the Infrared Data Association (IrDA) in
1993 [32]. The IrDA had been influenced by industry partners in defining
protocols and standards. One of the most challenging tasks was to increase the
data rate of the IR link.
1.2.1
System Structure
As discussed above, due to the physical properties of the link, most optical
wireless systems employed intensity modulation and direct detection (IM/DD).
Figure 1.1 showed a typical Infrared link using IM/DD [15]. () represents the
instantaneous optical power from the emitter, () indicates the instantaneous
current generated by the photodetector. Since the surface of the photodetector was
millions of square wavelengths at the received optical signal wavelength, the
optical link will not suffer from multipath fading effects that usually experienced
by the RF system [12].
The diffuse link, on the other hand, can provide more robustness for the optical
channel at a cost of reduced power and bandwidth efficiency. The transmitter and
receiver in a diffuse system established a connection by reflecting light from the
ceiling or other diffusely reflecting surfaces [15]. The users of a diffuse system
need not consider the alignment between transmitter and receiver. A constant
connection can be maintained, as long as the user was covered by the transmitter
signals illumination. The diffuse systems usually feature wide FOV receivers [61].
The first diffuse indoor diffuse Infrared wireless system was built by IBM in
1979, which achieved a data rate of 64kb/s and 125kb/s using phase shift keying
(PSK), and baseband pulse code modulation (PCM), respectively [10]. A 50Mb/s
diffuse link was achieved by researchers at Berkeley in 1994 which employed
OOK modulation and a decision feedback equaliser to mitigate the inter symbol
interference (ISI) [58].
According to orientation between the transmitter and receiver, the optical link can
also be divided into 3 categories: Directed, Hybrid and Nondirected. The directed
link refereed to the case when the transmitter and receivers were pointing in the
same direction in a LOS or a diffuse (Non-LOS) system. A hybrid link can
provide some degree of directionality but the receiver employed a wide angle
FOV to receive the optical signal. In the nondirected scenario, both transmitter
and receiver had a wide angle of FOV [15]. A detailed classification of optical
wireless links can be seen in Figure 1.2.
1.2.2
Optoelectronic Components
good choice for mass production and for quick adoption to the consumer market.
Table 1.3 listed detailed comparisons between LEDs and LDs.
Table 1.3 Comparison of LEDs and LDs (table adapted from [15, 21])
Characteristic
LED
LD
25 100 nm
Tens of kHz to Hundreds of
MHz
<10-5 5 nm
Tens of kHz to
Tens of GHz
Modulation Bandwidth
Special Circuitry Required
None
Eye Safety
Reliability
E/O Conversion Efficiency
Cost
High
10-20%
Low
Threshold and
Temperature
Compensation Circuitry
Must be Rendered Eye
Safe
Moderate
30-70%
Moderate to High
on the PPM, this including the multiple PPM (MPPM) [43], overlapping PPM
(OPPM), differential PPM (DPPM) [37], differential amplitude PPM (DAPPM)
[64], digital pulse interval modulation (DPIM) [65] and spectral efficient
modulation scheme such as the adaptively biased QAM (AB-QAM) [66] and 2Level 2-Pulse-Position Modulation (2L2PPM) [67] were reported.
The effects of ISI in diffuse links and the ambient light noise from background
illumination need to be considered when validating performance of optical
wireless systems [68, 69], which was usually ignored by most optical wireless
system researchers for model simplicity. Although techniques such as the use of
equalisation filters can be effective to reduce the ISI, yet were not optimised for
dynamic ISI interference, and usually came at cost of system complexity [12, 70].
11
The first chapter was the introduction, mainly providing the background and a
literature review on related topics. It also suggested the main problem to be solved
throughout the thesis, and discussed the possible solutions.
The second chapter concentrates on the channel models and channel interference.
Channel topologies together with artificial light model were discussed. Two major
types of noise source: the ISI caused by multipath dispersion and background
ambient light noise interference introduced by artificial light source were analysed.
Mathematic expressions and quantified noise parameters were discussed and
derived.
The third chapter began with the analysis of popular modulation schemes that had
been selected as candidate schemes for optical wireless communications. The
combined power and bandwidth properties, signalling structures and error
performance were covered. This also prepared the backgrounds for Chapter 4.
12
The fifth chapter addressed the application of fuzzy logic control concept for the
adaptive modulation. Followed by brief introductions of the artificial intelligence,
the control algorithms were explained. Example fuzzy inference models were
constructed. Adaptive neuro-fuzzy inference system (ANFIS) models were also
developed for the control pattern recognition.
The sixth chapter looked into the reliability issues of the optical channel. Adaptive
modulation schemes under combined channel interference were further
demonstrated by using the fuzzy logic control technique, the resulting system had
shown the capability of maintaining system stability under either or both of the
two types channel interferences induced by multipath ISI and background ambient
light.
The seventh chapter covered the conclusions and suggestions for future work.
Important results and methodology obtained from previous chapters were
summarised, and the possibilities for future directions were discussed.
13
Chapter 2
Channel Model
2.1
Introduction
2.2
Literature Review
2.2.1
2.2.1.1
Eye Safety
2.2.1.2
Classes of Lasers
2.2.2
Channel Topologies
2.2.3
Propagation Model
2.2.4
2.3
Channel Capacity
2.2.3.1
2.2.3.2
Channel Interferences
2.2.4.1
Multipath ISI
2.2.4.2
2.2.4.3
2.2.4.4
2.2.4.5
2.2.4.6
Problem Definitions
2.3.1
Main Challenges
2.3.2
Possible Solutions
2.4
Original Contributions
2.5
2.1 Introduction
The appropriate channel model for the optical wireless system depends on the
relative background optical noise levels where the system was deployed [12, 52,
14
72]. In the case of low background interference, the channel can be modelled by a
Poisson process. This was due to the random nature of the photons emitted from
the light source. When the background noise was high enough and comparable
with the optical signals, (in some cases this referred to optical signals other than
the source which operating at the same wavelength) the channel can be
approximately modelled by an additive white Gaussian noise (AWGN) model [15].
The exact channel model can be approached by combining both Poisson and
Gaussian distribution contributions. To obtain the combined formula, one key step
was to calculate the summation of Poisson and Gaussian stochastic variables. The
probability density expression for such a sum was easy to write down, but as it
contained an infinite summation, which made it numerically impractical [72].
Figure 2.1 Optical wireless system diagram (figure adapted from [21])
The equivalent channel model can be illustrated in Figure 2.2, where is the
input optical power, and () is the total photocurrent produced by the
15
Using the Gaussian model, the output current at the receiver () was given by:
= + ()
(2.1)
Where the symbol "" denotes convolution, since the optical signal was nonnegative and the average transmitted optical power must be constrained due to
eye and skin safety, so must satisfy the following:
0,
1
2
lim
()
(2.2)
These constraints greatly influenced the choice of signal design, channel model
and modulation selection. Note that in equation (2.2) the input represented
power, not amplitude. This was different to the conventional RF wireless channel,
where the power 2 () thus the mean square of the signal amplitude of the
channel input was limited. These unique constraints made the wireless Infrared
16
channel distinguished from the conventional linear Gaussian noise channel. The
resulting channel combines the filtered Gaussian noise characteristics of
conventional wire based channels with the IM/DD constraints of fibre-optic
systems [12]. Modulation schemes that were well suited to the conventional
channel may not be strong candidates for wireless optical channels. More details
on modulation will be discussed in Chapter 3.
2.2
Literature Review
2.2.1
Channel Capacity
The channel capacity was the highest rate in bits per channel use at which
information can be sent with arbitrarily low probability of error [29]. The capacity
of discrete-time memoryless channel subject to various input constraints had been
studied followed by Shannons information theory [73]. The most common input
constraints for the optical wireless channel were average power and bandwidth.
Since the early work of Gfeller into the optical wireless communication [10], the
capacity of the optical wireless communication channel had been an attractive
topic [12, 15, 74-76]. Recently a tighter higher and lower bound were reported for
the low signal to noise power (SNR) case [77]. Figure 2.3 showed the lower and
upper bounds together with L-PAM modulation with L= 2, 4 and 8, where ,
and were the original upper bound, updated upper and lower bound
respectively.
17
18
19
Table 2.1 Types of radiation and their likely effects on the human eye
(table adapted from [79])
Name
Ultra-Violet C
Ultra-Violet B
Ultra-Violet A
Visible
Wavelength
100 280 nm
280 315 nm
315 400 nm
400 760 nm
Eye Damage
Cornea
Cornea
Cornea & Lens
Cornea & Retina
Infra-Red A
760 nm 1.4 m
Infra-Red B
Infra-Red C
1.4 - 3.0 m
3.0 m - 1mm
Cornea
Cornea
880 nm
(infrared)
1310 nm
(infrared)
1550 nm
(infrared)
Class 1
Up to 0.2 mW
Up to 0.5 mW
Up to 8.8 mW
Up to 10 mW
Class 2
0.2 1 mW
N/A
N/A
N/A
Class 3A
1 5 mW
0.5 2.5 mW
8.8 4.5 mW
10 50 mW
Class 3B
5 500 mW
2.5 500 mW
45 500 mW
50 500 mW
Class I was the lowest class of laser and lasers in this class were believed to be
unable to cause eye damage even when shone directly into the eye for an extended
20
period of time. Class II lasers emited low-power, visible radiation that probably
cannot cause damage within 0.25 seconds if shone directly into the eye. Class III
lasers were those that can create a hazard in less than 0.25 seconds. These can
cause permanent damage to the naked eye. Class IV lasers had such high power
levels that they can create dangerous levels of radiation even after reflection from
dull surfaces [79].
2.2.2
Channel Topologies
The Lambertian model was usually adapted as the propagation model used to
model the wireless optical channel. The optical source (transmitter) can be
modelled by the following [12]:
+ 1
,
2
[ , ]
2 2
(2.3)
is defined as the optical power per unit solid angle originated from the
source with unit source orientation vector , is the total source optical power,
is the radiation lobe mode number, the source radiation pattern become more
21
directional when increases, this can be observed in Figure 2.5. is the angle
between incident path and normalised source orientation . Figure 2.5 showed a
typical relationship between source and the receiver. The source and receiver can
be denoted using their parameters = { , , ; }, = , , A , FOV .
and are optical source and receiving element respectively, Where and is
source and receiver position in three dimensional Cartesian coordinate with
[, , ]. is the normalised receiver orientation, A is detector effective surface
area, FOV is detector field of view.
2.2.3
Propagation Model
Figure 2.5 also showed the geometry set up of optical source, detector (receiver).
If the distance between transmitter and receiver was larger than the detector size,
22
; ,
=
For
cos
+ 1
( /)( /)
2
2
(2.4)
delta function, is speed of light in vacuum, R is the distance between the source
and receiver, = , is the angle between receiver and (
), is the angle between and ( ), where cos = (
)/ , cos
make sure only the incident light from transmitter that within receivers FOV were
counted for calculation, energy that fall outside the FOV will not contribute to the
total energy received by the detector, defined as, =
1 || 1
. In
0 || > 1
divided into a large set of small areas , refer to as reflector elements [82].
These areas were first considered as individual collecting elements as indicated in
previous section, and optical power received can be obtained using the source and
detector model. Each element was then act as a point source that re-emits the
collected signal scaled by the surface reflection coefficient ( 1 ). By
summarising each of the reflector elements, the one reflection impulse response
can be expressed as [82]:
; ,
+1
=1 2 2 2
cos ( , , ) (
( + )
) (2.5)
24
25
+ 1
; , =
2
=1
, 90o
; , )
(2.6)
(; , ) is
26
Figure 2.8 showed a multi-path data link when LOS is available. In this case, the
multipath contribution distorted the received optical pulse as late arrived pulses
also contributed to the detected optical power at the receiver.
27
Figure 2.9 showed that an office separator can block most of the transmitted IR
signals. The receiver can only communicate with the transmitter through a multipath link. In this case, the multi-path links can cover areas that cannot be reached
through a LOS links.
28
100
k=1
k=2
k=3
90
80
70
60
50
40
30
20
10
0
10
20
30
40
50
Time (ns)
60
70
80
90
100
k=1
k=2
k=3
90
80
70
60
50
40
30
20
10
0
10
20
30
40
50
Time (ns)
60
70
80
90
29
From Figure 2.10, it can be demonstrated that as number of bounces increase, the
received optical power decreased significantly after the reflection from room
surface. Since there was no separator between the transmitter and receiver, optical
wireless communication systems in room A relied on signals with lower reflection
orders (e.g. k=1). Although contributions from second and third reflections count
towards the total received optical energy at the receiver, compared to first order
reflection, they were not significant. It was a totally different scenario for room B.
In Figure 2.11, by placing a separator between transmitter and receiver, most of
the first order reflection energy was blocked. A more reflective floor also helped
shifting the received energy to the second order reflection. Thus for room B,
communication systems can establish connections using second order reflections.
Appling the same system in room A to room B would results in substantial system
degradation if system parameters remain the same. This was because the
contribution of the received optical energy had been shifted due to the blockage.
From above two figures, it had been demonstrated that channel impulse response
can be significantly different even with the same geometry (e.g. size of two rooms
were same) and transmitter-receiver locations. This suggested that the impact of
multipath reflection cannot be neglected when validating modulation schemes.
Channel dynamics need to be considered when designing optical communication
systems.
For the multiple reflection model, in order to get more accurate approximation of
the impulse response (), the reflection orders was preferred to count as many
reflections as possible, while the time needed to calculate () also increases
30
exponentially with [12], even with latest computers, the calculation time was
still considerably long for higher order reflections.
This model can be expressed by the path loss and delay spread [53]:-
66
+
13
12 11
(2.7)
(2.8)
31
0, < 0
, depends on the relative location of the transmitter and
1, > 0
receiver, when the transmitter and receiver were collocated, = 2/, where
is the ceiling height , is speed of light, is the root mean square (RMS)
delay spread. In this thesis, it was assumed that the transmitter and receiver were
collocated, as discussions will not loss generality with this assumption regarding
to non collocated cases. From equation (2.7), the channel impulse response can be
quantified by the ceiling height, thus can be used to reflect severity of the ISI
caused by multipath propagation.
32
Figure 2.13 showed the background radiation power spectral density of sunlight,
incandescent and florescent lighting. The Si-Photodiode responsivity was also
indicated with dotted lines. This showed the Infrared optical channel can suffer
intense distortion caused by the background ambient noise. The sunlight and
incandescent light exhibited less periodic characteristics than the florescent light.
Thus an optical filter can be used to effectively block much of these two types of
radiation. The florescent lamps can be grouped into two categories: lamps driven
by conventional ballast and electronic ballast (also known as the energy saving
lamp). The latter became more popular as the energy saving feature. The
incandescent and florescent lamps exhibited different spectrum. The artificial
lamp radiation pattern can be found in following Figure 2.14 [84].
33
(a)
(b)
(c)
Figure 2.14 Typical artificial light interference time waveform and spectrum
of (a) Incandescent lamp (b) Fluorescent lamps driven by conventional
ballast and (c) Fluorescent lamp driven by electronic
ballast (energy saving lamp) (figure adapted from [84])
It can be observed from Figure 2.14 that radiation from electronic ballast driven
florescent lamps had a stronger periodic nature.
34
=
+
1
20
2 100 50 + + 2100 +
=1
+
0 cos 2 + 0 +
2
11
22 + (2.9)
=1
= 10(13.1 ln
10050 +27.1)/20
= 10(20.8 ln
100 +92.4)/20
, 1 20
, 1 20
(2.10)
(2.11)
35
1
2
3
4
5
6
7
8
9
10
4.65
2.86
5.43
3.9
2
5.98
2.38
4.35
5.87
0.7
0
0.08
6
5.31
2.27
5.7
2.07
3.44
5.01
6.01
11
12
13
14
15
16
17
18
19
20
1.26
1.29
1.28
0.63
6.06
5.49
4.45
3.24
2.07
0.87
6
6.17
5.69
5.37
4
3.69
1.86
1.38
5.91
4.88
Table 2.4 Typical value for amplitude and phase parameters of high frequency
components (table adapted from [85])
0
1
2
3
4
5
(dB)
-22.22
0
-11.5
-30
-33.9
-35.3
(rad)
5.09
0
2.37
5.86
2.04
2.75
6
7
8
9
10
11
(dB)
-39.3
-42.7
-46.4
-48.1
-53.1
-54.9
(rad)
3.55
4.15
1.64
4.51
3.55
1.78
A sample waveform of the interference signal with =1A/W and =1W can be
obtained using equation (2.9) and demonstrated in the following Figure 2.15 [68].
36
37
(a)
(b)
(c)
Figure 2.16 Modulation performance in channel limited by (a) shot noise only
(b) incandescent light without HPF (c) incandescent light with HPF
(figure adapted from [31])
In Figure 2.16, comparing (b) with (a), the incandescent light interferences
resulted 24dB penalty for OOK, 16dB for 16-PPM with threshold detection (TH)
and 1.5dB for 16-PPM with maximum-a-posterior (MAP) detection for a 1Mbps
data link. Comparing (c) with (b), applying HPF can effectively reduce the
interference caused by incandescent lamp for both OOK and 16-PPM modulation
schemes. Modulation performance comparison under fluorescent light interference
can be found in following Figure 2.17.
38
(a)
(b)
Apart from HPF, better BER performance can also be obtained by designing the
system modulation/demodulation to achieve a higher average BER first, and then
reducing the results to the target BER value through error correction codes (ECC),
in conjunction with interleaving [86]. However, ECC method often resulting in
reduced transmission data rate [87].
39
2.3
Problem Definitions
2.3.1
Main Challenges
From previous discussions, the ISI caused by multipath propagation and artificial
light interference from fluorescent lamp driven by electronic ballast were two
major interferences, and these need to be taken into account when validating
modulation schemes. The severity of multipath ISI can be quantified by the
distance variable in the ceiling bounce model. HPFs were effective for mitigating
interference induced by incandescent light and conventional-ballast-driven
fluorescent light but not for the electronic-ballast-driven fluorescent light. The LPPM modulation scheme presented a good candidate under severe interferences
caused by artificial lighting. Yet as were discussed in Chapter 3, the L-PPM
modulation scheme was not bandwidth efficient compared to L-PAM and OOK
schemes.
The main challenge faced by this thesis was to seek the most optimised
modulation scheme that can provide maximum system throughput while capable
of withstanding most if not all of the intense channel interferences at a target BER
requirement. This defined a dilemmatic situation, modulation schemes such as the
L-PPM proved to be less susceptible to artificial lighting interferences but not
bandwidth efficient. Bandwidth efficient schemes such as the OOK and L-PAM
were prone to artificial lighting interferences. This led to a natural conclusion of a
modulation scheme that can combine benefits from both above candidates and
able to avoid the drawbacks of each individual scheme. The multilevel pulse
amplitude and position modulation (PAPM) thus been selected as the new
40
2.3.2
Possible Solutions
Similar modulation combinations had been proposed in the literature, such as the
differential amplitude pulse position modulation (DAPPM) [64], which combined
the differential pulse position modulation (DPPM) and pulse amplitude
modulation (PAM). Multilevel digital pulse interval modulation (MDPIM)
combined dual header pulse interval modulation (DH-PIM) with PAM [88]. Both
DAPPM and MDPIM can increase data throughput due to the PAM element while
enjoy the benefits from DAPPM and DPIM elements, such as the inherent symbol
synchronisation capability and improved transmission rate and bandwidth
requirements. With many new PPM derivatives being reported, the L-PPM
scheme still remain attractive for its power efficiency and improved immunity to
the fluorescent lamp induced noise [15]. The 4-PPM modulation scheme was
adopted by the IrDA in its physical layer specification [89].
In order to compare and validate the PAPM modulation under different types of
interferences, detailed analytical model together with BER, SNR and data rate
relationships were needed. Wong et al [68] had developed an analytical model for
studying multipath ISI and electronic-ballast-driven fluorescent light interferences.
Yet Wongs model was limited to OOK, 2-PPM and sequence inversion keying
(SIK) direct sequence spread spectrum, and the multipath ISI considered was only
valid for a specific room set up. Moreira et al [31] developed mathematical
models for analysing the artificial light interference for OOK and L-PPM of
41
Appropriate control modules needed to realise the dynamic adaptations for the
proposed modulations. This can be facilitated by employing artificial intelligence
algorisms. Simulation results needed to be compared with analytical discussions.
2.4
Original Contributions
The contributions presented within this thesis can be summarised into three
constituent parts:
42
3. The adaptive modulation concept developed in this thesis provided some insight
on the stabilising issues of high speed OW communication link. In Chapter 6, by
adaptive modulation parameters optimisation, system throughput can be improved
compared to non-optimising case.
2.5
Summary
The Infrared communication channel can be characterised by LOS and diffuse
prorogation model. Channel noise mainly came from background noise and
multipath ISI. The achievable data rate of a channel was restricted to the available
bandwidth that a specific channel can provide. The impulse response of the
channel was depended on transmitter, receiver location and orientation, dimension
of the room where the system was deployed. Eye safety regulations defined the
maximum allowed average and peak optical power that can be used in an optical
wireless link.
43
Conclusions
The unique characteristics of the optical wireless channel exhibited challenges and
opportunities. Constrains and interferences presented to the channel need to be
taken into account when designing communication systems. In order to improve
channel throughput, the first step was to set up the appropriate channel model.
This included fully understanding the mathematical model of the channel, noise
sources and error performance under each or combined interferences. Partially
represented channel model cannot be used for validating system performances.
Channel behaviours can be described for a specific scenario. Channel frequency
response can vary significantly according to transmitter and receiver location.
Furthermore, analytical models developed for the optical wireless channel can
only be applied to validate the modulation scheme performance when given the
exact channel parameters.
In order to improve the channel throughput under the presence of channel limits,
next chapter considered different modulation schemes proposed in the literature,
and their performance under the constraints imposed by the challenges in
designing a robust optical wireless system.
44
Chapter 3
3.1
Introduction
3.2
Modulation Schemes
3.3
3.4
3.1
3.2.1
On-Off-Keying (OOK)
3.2.2
3.2.3
3.2.4
OOK
3.3.2
PAM
3.3.3
Introduction
The optical channel is quite different from the conventional RF channel. This
consequently resulted in a different approach when it came to the modulation
design. Modulation schemes which fit well in electromagnetic channels were not
necessarily perform well in the optical domain [12]. Modulation techniques
remained an active topics amongst both academic researchers and industrial
communication system engineers [16, 37-40, 42, 43, 91]. Depending on the nature
45
Figure 3.1 Family tree of pulse modulation schemes for optical wireless
systems (figure adapted from [10, 36])
3.2
Modulation Schemes
Figure 3.2 Comparison of (a) NRZ-OOK pulse (b) RZ-OOK pulse with
duty cycle = 0.5 (figure adapted from [15])
47
p (t )
for
t [0, T )
(3.1)
0
elsewhere
The power spectral density (PSD) of OOK can be calculated using the following
equation [21], and its PSD curve can be found in Figure 3.3:
= 2 + 2 2 ( )
(3.2)
Where the first and second part of equation (3.2) are the discrete and continuous
portions respectively, is the Dirac delta function, is the average
transmitted optical power, is the symbol interval, and =
sin
()
48
The bandwidth required by OOK is = 1/, the inverse of the pulse width, its
bit error rate (BER) is [12]:
(3.3)
where 0 is the power spectral density of the white Gaussian noise and is
the customary Q-function of digital telecommunications. is the average
received optical power. Since x , and is monotonically decreasing, the
inverse 1 () where x {0,1} is straightforward to obtain [93]. The power
requirement for OOK is [12]:
POOK N 0 Rb Q 1 ( BEROOK )
(3.4)
49
B L PAM
Rb
1
BOOK
log 2 LPAM log 2 LPAM
P
log 2 LPAM
BERL PAM Q L PAM
N 0 Rb
LPAM 1
PL PAM
LPAM 1
log 2 LPAM
N 0 R0 Q 1 ( BERL PAM )
(3.5)
(3.6)
(3.7)
50
To compare with the OOK system, when achieving the same BER:
(3.8)
LPAM 1
PL PAM
log 2 LPAM
POOK
(3.9)
The above equation is under the assumptions of a high Signal to Noise Ratio
(SNR), moderate values of ( 2), and a given BER.
PL PPM
B L PPM
LPPM
2
POOK
log 2 LPPM
LPPM
BOOK
log 2 LPPM
(3.10)
(3.11)
51
BM n PAPM
n
BOOK
log 2 nM
(3.12)
52
PM n PAPM
2M 2
POOK
n log 2 nM
(3.13)
3.3
In this section, the BER expression for different modulation schemes under both
ISI and background ambient light noise can be derived. During an instance of time
interval , energy of kth sample arriving at the threshold detector is [68]:
= + ,
=+
(3.14)
0 /2, is contribution of
53
= =
(3.15)
1 =
6
+1
(3.16)
To calculate the ambient light interference, the channel impulse response within a
time interval were considered, assume at time after channel inference, kth pulse
54
(+1)
1 () = ( + 1)7
+ 1
(3.17)
Similar to Wongs method [68], in order to consider the impact of ambient light
interference, the ambient light noise to signal power ratio (ASR) can be used to
indicate the degree of ambient light interference. The ratio can be integrated into
equation (3.14) by dividing , express y, z, v and in units of :
= /
= + ,
=
=+
= , = =
(3.18)
0 =
(3.19)
0 /2 and =
= 0 = 2 /0
(3.20)
55
3.3.1 OOK
For given threshold , the probability of with bit 0 = 1 to fall below threshold
is
1
0
= =
(3.21)
=0
( ( )), where
= 1 =
(3.22)
=1
56
the ambient light interference. The probability can be expressed by the following
[68]
1
=
1
2
1 1
0 + 1
0 =0
0 =1
(3.23)
Where is the considered time period of ambient light interference,
1
2
is the
_ = 1
(3.24)
3.3.2 PAM
For PAM case, thresholds can be used to make decision on received signal
power. There were three possible events:
1. Detection success.
2. Over detection failure (over threshold to next level)
3. Under detection failure (not enough power to current level)
57
1
=
0 =0
0 =
,0 +
0 =
, + , +
0<<
(3.25)
1
=
1
+ 1
0 =0
0<<
+ 1
0<<
+ 1 ] (3.26)
0 =
1
=0
is the
convolved signal after the optical wireless channel. In fact, OOK can be treated as
2-PAM, as 2-PAM had two levels of amplitude change, if assume the possible
levels were 0 and 1, then OOK can be included into L-PAM modulation
schemes.
58
= = (1 , 2 , 3 , , , , )}
(3.27)
Where = , is the last chip of the received pulse. The probability of all
outcomes of the detection event can be expressed as following
= 1
(3.28)
= 1 1 2 2
(3.29)
1
2
2
2
where =
variance. Successful detection of m-th amplitude level of the primary symbol was
a condition of following two events:
59
< = +
(3.30)
< < +
(3.31)
+ =
,
+,
<
=
1 ,
,
>0
=0
(3.32)
1
2
= 1
(3.33)
Detailed derivation can be found in Appendix III-1, the probability of success for
both equation (3.30) and equation (3.31) is then
Where
(3.34)
yields:
(3.35)
60
The full probability of correct detection over all possible chips sequences and
1
( )(
+1)
3.4
1
= 1 =
(3.36)
Summary
Modulation schemes preferred for the optical wireless channel were introduced.
The combined power and bandwidth efficiency expressions were listed. The
detection error probability of the three baseband modulation schemes were listed
and derived for optical channel impaired by both ISI and background ambient
light noises. The obtained analytical model can be used to count for any
modulation order of PAM, PPM and PAPM modulation schemes. This extended
the previous mathematical model and provided a useful platform to validate
modulation schemes under single or multiple interferences.
61
Chapter 4
Adaptive Modulation
4.1
4.2
4.3
4.4
4.1
Introduction
4.1.1
Channel Model
4.1.2
Adaptive Modulation
4.2.1
Adaptive L-PAM
4.2.2
Adaptive L-PPM
4.2.3
Adaptive M-n-PAPM
4.3.2
Introduction
off between the bandwidth and power efficiency was non avoidable, modulation
schemes can adaptively tune amplitude levels or pulse positions in order to
maintain the maximum possible throughput under interferences [95].
Wong et al analysed ISI and ambient noise impact for different modulation
techniques under specific channel geometry set ups (a room size of 5m5m3m)
[68]. Their discussions were limited to include comparison amongst OOK, 2-PPM
and SIK only. In this chapter, the performance of popular modulation schemes
were discussed under a more general channel model, e.g. not limited to a specific
room set up, with modulation schemes extended to include L-PAM, L-PPM and
63
M-n-PAPM. In terms of combating the ISI and background ambient noise, the
proposed adaptive modulation scheme were analysed under different channel
impairments. Simulation results were used to validate the performance of the
proposed scheme with other candidate for the wireless optical communication
channel.
The adaptive modulation scheme here was initially intended to mitigate the data
rate drop in a diffuse optical link, where multipath distortion was present [71].
However, the adaptive modulation was not limited to a diffuse model. Modulation
techniques developed in this chapter were also suitable for LOS systems, where
multipath distortion was not regarded as significant compared to that of diffuse
systems, since the interference from ambient light noise can be reduced by
increasing the optical pulse intensity, thus increasing the SNR. In the interests of
data rate recovery, the optimum modulation scheme parameter under different
system degradations can be obtained through searching algorisms. The candidate
modulation schemes had been chosen, based on the merit of combined power and
bandwidth efficiency, as detailed in Chapter 3. For model simplicity, the
following assumptions were made:
64
The impulse response of the ceiling bounce model can be plotted versus time at a
given ceiling height H, in Figure 4.1, = 10, time step was 1.
In Figure 4.1, the energy of the received optical pulse decreases with time, while
most energy (e.g. 90%) arrived within 30 in this case, with delayed tails lasting
up over to 70 . The impulse response was directly related to parameter
according to equation (2.7), and the relationships of the impulse response under
different ceiling height can be obtained in Figure 4.2.
65
The standard system model was derived from the OOK modulation scheme. First,
for given link parameters, a corresponding normalised data rate compensation
ratio was derived from the OOK scheme. Second, a multilevel modulation scheme
performed a search within its available system status to find its data rate
compensation ratios. Finally, comparing these ratios to the normalised OOK ratio,
66
the system status with best approximation to the normalised ratio represented the
optimum candidate for the adaptive modulation scheme.
Thus, in the test model two sets of BER test data were used. The first set ranged
from 10-9 to 10-7, to represent moderate degradation and second set ranged from
10-7 to 10-4, to represent severe degradation. The data rate was chosen at 4Mbps
and 250Mbps.
67
4.2
Adaptive Modulation
The adaptive modulation scheme was proposed to improve data throughput when
the optical wireless channel was distorted due to ISI and intense background
ambient light noise [71]. By changing the optical pulse amplitude levels or
positions, signal pulse energy and bandwidth requirement can be adjusted. This
resulted changes in SNR and data rate respectively, which in turn provides BER
improvements at the cost of power consumption or reduced data rate. This
approach was different from other techniques, such as spread spectrum, which
increased system complexity. Adaptive modulation schemes were especially
suitable
for
high
speed
indoor
wireless
optical
environment,
where
communication systems operate under intense background ambient light noise and
ISI interference.
The OOK modulation scheme was again treated as a reference when comparing
with other schemes. The power compensation ratio was obtained from the BER
variation ratio. For moderate degradation, with initial BER=10-9 and final
BER=10-7, from equation (3.4) the power compensation ratio is given by:
68
POOK _ i
POOK _ f
N 0 ROOK _ i Q 1 ( BEROOK _ i )
N 0 ROOK _ f Q 1 ( BEROOK _ f )
ROOK _ i Q 1 ( BEROOK _ i )
ROOK _ f Q 1 ( BEROOK _ f )
(4.1)
Where POOK_i and ROOK_i are the initial power requirement and data rate to achieve
initial BEROOK_i. POOK_f and ROOK_f are the power requirement and data rate to
achieve the varied final BEROOK_f, and in the case of constant average power,
POOK_i = POOK_f, and equation (4.1) becomes:
ROOK _ i
ROOK _ f
Q 1 ( BEROOK _ f )
1
Q ( BEROOK _ i )
(4.2)
Note equation (4.1) and equation (4.2) were only valid when satisfying the
following conditions:
ROOK _ i 0
ROOK _ f 0
(4.3)
Q 1 ( BEROOK _ i ) 0
1
Q ( BEROOK _ f ) 0
(4.4)
(4.5)
69
Final Rb (Mbps)
5.3
332.7
Severe Degradation
Initial Rb (Mbps)
4
250
Final Rb (Mbps)
7.8
488.6
In Table 4.1, with moderate degradation, the data rates changed from 4Mbps and
250Mbps to 5Mbps and 333Mbps respectively, which was 33.1% of throughput
variation. For severe degradation, the variation was 95.5%
Comparing data rate variation for systems operating at 4Mbps, the difference
between the moderate and severe model was not significant. By contrast, for a
system operating at 250Mbps, the data rate loss between the moderate and severe
model was large. This suggested that, at higher speed, the OOK modulation
scheme was more susceptible to BER variation.
4.2.1
Adaptive L-PAM
The L-PAM modulation required more power to achieve the same level of BER as
L-PPM. Thus L-PAM modulation was not preferred in terms of power efficiency
[15]. While considering the bandwidth requirements, L-PAM modulation was
more efficient than L-PPM, as, by assigning different levels of amplitude to
70
represent a symbol sequence, the L-PAM signal cannot span its symbol sequence
along the time axis, which was the case for L-PPM. This can be demonstrated in
Figure 4.3. Higher level PAM was throughput-efficient scheme, as it allowed
more data to be transmitted compared to PPM. Detailed procedures and Matlab
program for Figure 4.3 can be found in Appendix IV-2.
71
From equation (3.7), the average power requirement of L-PAM can be found for a
given BER, following equation (4.1); the initial and final state of the L-PAM
scheme can be described as:
LPAM _ i 1
PL PAM _ i
PL PAM _ f
log 2 LPAM _ i
LPAM _ f 1
log 2 LPAM _ f
Where PL-PAM_i, RL-PAM_i and LPAM_i indicate the average power requirement, the
data rate and amplitude level to achieve BERL-PAM_i. PL-PAM_f , RL-PAM_f and LPAM_f
represent the average power requirement, data rate and amplitude level
respectively to achieve BERL-PAM_f . Following the same discussion as for OOK,
the initial error rate, the varied error rate, and the initial data rate satisfy the
following:
(4.7)
Keeping the average power requirement constant, equation (4.6) can yield:
PL PAM _ i
PL PAM _ f
72
RL PAM _ i
RL PAM _ f
LPAM _ f 1
log 2 LPAM _ f
LPAM _ i 1
log 2 LPAM _ i
1
Q ( BERL PAM _ f )
Q 1 ( BER
L PAM _ i )
(4.8)
From equation (4.8), the data rate ratio was a function of amplitude levels and
BER, where the first item in equation (4.8) can be used as a ratio to balance the
degradation caused by the variation of the BER.
Comparing equation (4.8) with equation (4.2), the L-PAM scheme provides a ratio
factor by changing the amplitude level LPAM. This ratio factor was a function of
LPAM. The set formed by LPAM was a subset of the natural number set N with a
condition {LPAM | LPAM 2, LPAM N } . By selectively choosing values of LPAM_i
and LPAM_f, the ratio factor in equation (4.8) can compensate for the data rate
reduction caused by the BER variation.
73
log 2 LPAM _ f
LPAM _ i 1
log 2 LPAM _ i
LPAM_f
LPAM_i
0.4
0.2
2.5
0.6
4.5
1.8
In Table 4.2, the adaptive factor matrix was obtained by changing pulse amplitude
levels. The maximum value is 4.5, which was provided by changing the amplitude
level from 2 to 4. The minimum value was 0.2, and obtained by changing the
amplitude from 4 back to 2. This suggested that a reverse level change will not
necessary give the same adaptive ratio. The ratio table was symmetric along the
table axis.
For moderate system degradation, which the initial BER=10-9 and the final
BER=10-7, substituting BER values into (4.8), the degradation factor is thus:
Q 1 ( BERL PAM _ f )
Q 1 (10 7 )
1
1 9 0.8
Q ( BERL PAM _ i )
Q (10 )
2
(4.9)
Considering this degradation factor, together with the candidate values in Table
4.2, in order to find the optimum adaptive level factor, the inverse of above ratio
was compared with every value in Table 4.2. The preferred value made the right
74
hand side of equation (4.8) approach 1. Thus, subtracting the inverse value of the
above ratio, and applying it to every element in Table 4.2, gives Table 4.3
LPAM_f
LPAM_i
-1
2
-0.2
-0.7
-0.8
0.9
-0.2
-0.6
2.4
0.3
-0.2
From Table 4.3, the values along the diagonal were self level comparison, and
thus were ignored. Positive values indicated the situation where the adaptive
factor was greater than the interference factor, and suggested that the adaptive
factor was sufficient to compensate the degradation caused by the BER variation.
The negative situations were vice versa. Note the values in Table 4.3 also
indicated the quality of the adaptive factors. These values reflected how close the
adaptive factors can approach the interference factor. Thus, the absolute values
were taken when processing comparisons. The smaller the values, the better the
adaptive abilities become. This can be demonstrated in the following Figure 4.4.
75
2.5
Initial L-PAMi = 2
Initial L-PAMi = 3
Initial L-PAMi = 4
1.5
0.5
-0.5
-1
3
Final L-PAMf Values
RL PAM _ i
RL PAM _ f
LPAM _ f 1
log 2 LPAM _ f
LPAM _ i 1
log 2 LPAM _ i
RL PAM _ f RL PAM _ i
1
Q ( BERL PAM _ f )
Q 1 ( BER
L PAM _ i )
LPAM _ i 1
log 2 LPAM _ i
L
1
PAM _ f
log 2 LPAM _ f
1
Q ( BERL PAM _ i )
(4.10)
Q 1 ( BER
L PAM _ f )
76
Final Rb (Mbps)
5.3
325
Initial Rb (Mbps)
4
250
Final Rb (Mbps)
3.6
225
From Table 4.4, the adaptive L-PAM performed similarly to the OOK scheme
under moderate BER degradations and lower data rate. However, the adaptive LPAM outperformed OOK under severe BER degradations, especially when
operating at a higher data rate. Thus adaptive L-PAM modulation schemes with
LPAM {2,3,4} can provide data rate recovery compared to OOK. The adaptive LPAM modulation schemes indicated an improvement of data rate recovery under
severe degradation. This resulted in a 30% throughput variation for the 4Mbps
link and 90% for the higher 250Mbps. Compared to OOK, L-PAM reduced the
variation range.
4.2.2
Adaptive L-PPM
In Figure 4.5, it was seen that by changing from 2-PPM to 16-PPM, the power
requirement reduced 7.5dB. This came at a cost of twice the bandwidth
requirement. The power to bandwidth ratio for L-PPM was then 7.5 / (4-2) = 3.75
dB per bandwidth unit. This was 3 times more power efficient than PAM.
The L-PPM modulation scheme was a power efficient scheme, and 4-PPM
modulation was adopted in the IrDA standard for 4Mbps data links [89]. In
contrast with the L-PAM scheme, under moderate pulse position levels
(e.g. > 2), the L-PPM scheme only required half, or even less of the power
required by other schemes such as the OOK and PAM. The L-PPM modulation
78
scheme was not bandwidth efficient [15]. Higher order pulse positions can lead to
higher bandwidth consumption.
Next, the data rate recovery abilities for L-PPM were examined. In a similar way
to the adaptive L-PAM, the OOK modulation scheme was treated as a benchmark.
According to equation (3.10) the average power requirement of L-PPM for a
given BER can be represented by relating to OOK as follows:-
PL PPM _ i
PL PPM _ f
LPPM _ i
LPPM _ f
2
N 0 RL PPM _ i Q 1 ( BERL PPM _ i )
log 2 LPPM _ i
2
N 0 RL PPM _ f Q 1 ( BERL PPM _ f )
log 2 LPPM _ f
(4.11)
where PL-PPM_i, RL-PPM_i and LPPM_i indicate the initial average power requirement,
data rate and pulse position levels required to achieve BERL-PPM_i. PL-PPM_f, RLPPM_f
and LPPM_f represent the average power requirement, data rate and pulse
(4.12)
Since the average power requirement was fixed, equation (4.11) can be rearranged
to represent data rate:
79
RL PPM _ i
RL PPM _ f
1
LPPM _ f log 2 LPPM _ f Q ( BERL PPM _ i )
(4.13)
The data rate ratio can be represented by two parts, the adaptive part formed by
pulse position levels and variation of system BER. Comparing equation (4.13)
with equation (4.8) and equation (4.2), and similar to the adaptive L-PAM, the
adaptive L-PPM scheme included the pulse position level LPPM as an adaptive
factor.
This
ratio
factor
was
function
of
LPPM
and
it
satisfied
L
PPM _ f log 2 LPPM _ f
LPPM_f
LPPM_i
2
0.4
0.3
2.4
0.6
1.7
In Table 4.5, the maximum ratio provided by this matrix was 4, which was
obtained by shifting L from 2 to 4. The minimum ratio was 0.3 by changing pulse
positions from 4 back to 2. Similar to PAM, the adaptive ratio was symmetric
along the matrix diagonal. Under moderate system degradation, substituting
BERL-PPM_i=10-9 and BERL-PPM_f =10-7 into equation (4.13), given the degradation
80
LPPM_f
LPPM_i
-1
2
-0.25
-0.68
-0.81
0.79
-0.25
-0.55
2.01
0.26
-0.25
Similar to Table 4.3, in Table 4.6, the values along the diagonal were self level
comparison, and thus were discarded. Positive values indicated the adaptive factor
was sufficient to compensate the degradation caused by the BER variation. The
negative values indicated the reverse situation. This can be further demonstrated
by Figure 4.6, where the optimum ratio can be compared to other values.
81
2.5
Initial L-PPMi = 2
Initial L-PPMi = 3
Initial L-PPMi = 4
1.5
0.5
-0.5
-1
3
Final L-PPMf Values
From Figure 4.6, the optimum value can be achieved by changing pulse positions
from 3-level PPM to 4-level PPM. Note this was similar to the L-PAM model. For
moderate degradation, LPPM_i=3, LPPM_f=4, BERL-PPM_i=10-9 and BERL-PPM_f=10-7
of RL-PPM_i=4Mbps and RL-PPM_i=250Mbps must be substituted into equation (4.13)
respectively. For severe degradation, BER was replaced with BERL-PPM_i=10-7 and
BERL-PPM_i=10-4, so the data rate recovery for adaptive L-PPM can be obtained.
Adaptive L-PPM schemes had an improved performance over adaptive L-PAM,
Referring to Table 4.1 for OOK and Table 4.4 for adaptive L-PAM, the data rate
recovery of adaptive L-PPM can be represented in the following Table 4.7.
82
Final Rb (Mbps)
5.0
315
Initial Rb (Mbps)
4
250
Final Rb (Mbps)
3.4
215
4.2.3
Adaptive M-n-PAPM
Now considering adaptive PAPM, similar to discussions for adaptive PAM and
PPM, the adaptive PAPM can realise adaptation by changing both amplitude and
number of pulse slots. The normalised power and bandwidth requirements can be
found in Figure 4.7. Detailed procedures and Matlab program for Figure 4.7 can
be found in Appendix IV-4.
83
In this section, adaptive M-n-PAPM was discussed using the same models which
were applied to OOK, PAM and PPM. The nomenclatures were the same as used
in Chapter 3. M represents the amplitude levels and n represents the number of
84
PM n PAPM _ i
PM n PAM _ f
Mi 1
N 0 RM n PAPM _ i Q 1 ( BERM n PAPM _ i )
ni log 2 ( M i ni )
(4.14)
M f 1
1
N 0 RM n PAPM _ f Q ( BERM n PAPM _ f )
n f log 2 ( M f n f )
(4.15)
Since system operated under constant average power constraint, equation (4.14)
can be rearranged to relate data rate with amplitude, pulse position and error rate:
85
RM n PAPM _ i
RM n PAPM _ f
( M f 1) 2
2
(4.16)
Q 1 ( BER
( M i 1) 2
M n PAPM _ i )
ni log 2 ( M i ni )
Comparing equation (4.16) with other modulations, the adaptive factor of M-nPAPM was a function of both M and n. This formed the basis for the adaptive Mn-PAPM. The efficiency of the modulation scheme depended on the values
selected. Similar to PAM and PPM, the data rate ratio can be represented by a
function of amplitude levels and pulse position numbers, together with channel
BER in equation (4.16). Based on the formulas obtained, the power and
bandwidth efficiency of the adaptive M-n-PAPM can be observed by applying the
experimental models. As M-n-PAPM contains two variables, the system states
table was thus included more candidate levels. This certainly benefited the
optimum value searching process.
Similar to adaptive PAM and PPM, the ratio factor can also be treated as a
function of M and n. The resulting set satisfies {M , n | M , n 2M , n N} .
Following the above conditions; a joint level ratio table can be organised by
selectively choose value of Mi, ni and Mf, nf. The obtained ratio factor was then
applied to the model discussed earlier to test the ability of data rate recovery
against channel degradation. The adaptive ratio factor can then be determined, as
in Table 4.8
86
( M _ i 1) 2
n_ i log 2 ( M _ i n_ i )
( M _ f 1) 2
n_ f log 2 ( M _ f n_ f )
M_i = 2
M_i = 3
M_i = 4
M_f = 2
M_f = 3
M_f = 4
n_f = 2
n_f = 3
n_f = 4
n_f = 2
n_f = 3
n_f = 4
n_f = 2
n_f = 3
n_f = 4
n_i = 2
1.9387
0.72702
1.3373
2.0165
0.54
0.96794
1.44
n_i = 3
0.5158
1.5474
0.375
0.68979
1.0401
0.27853
0.49927
0.74276
n_i = 4
0.33333
0.64624
0.24234
0.44577
0.67218
0.18
0.32265
0.48
n_i = 2
1.3755
2.6667
4.1264
1.8394
2.7737
0.74276
1.3314
1.9807
n_i = 3
0.74777
1.4497
2.2433
0.54364
1.5079
0.4038
0.7238
1.0768
n_i = 4
0.4959
0.96141
1.4877
0.36053
0.66317
0.26779
0.48
0.71409
n_i = 2
1.8519
3.5902
5.5556
1.3463
2.4765
3.7343
1.7925
2.6667
n_i = 3
1.0331
2.0029
3.0994
0.7511
1.3816
2.0833
0.55789
1.4877
n_i = 4
0.69444
1.3463
2.0833
0.50488
0.92869
1.4004
0.375
0.67218
87
From Table 4.8, the obtained adaptive factor table for adaptive M-n-PAPM
contained more values than other modulation schemes. It was 9 times larger than
that of both L-PAM and L-PPM. This certainly resulted in a better precision than
the other two modulation schemes. Comparing Table 4.2 and Table 4.5, Table 4.8
exhibits some similar properties, e.g. values along diagonal were equal to 1.
Similarly to previous discussions, the optimum adaptive factor made the best
possible match for equation (4.16), i.e. it can choose adaptive ratio factors to make
the initial and final data rate compatible, thus maximising the data rate
compensation. The comparable basis depended on the above BER ratio. The
optimum adaptive ratio can be obtained by two steps: Firstly, subtraction of the
inverse BER ratio from every element in Table 4.8 was done. Secondly, a
minimum search was performed within the results. The minimum data reflected
the most optimum system level under the given BER condition. Thus the optimum
system level can be identified using this algorithm. Note that values along the
diagonal were not considered as valid, as these were self comparison. Applying
the above calculation to Table 4.8 results in the following:
88
Table 4.9 Table 4.9 Comparison of adaptive and interference ratio for M-n-PAM
Final Value of M_f and n_f
RM n PAPM _ i
RM n PAPM _ f
M_i = 2
M_i = 3
M_i = 4
-1
M_f = 2
M_f = 3
M_f = 4
n_f = 2
n_f = 3
n_f = 4
n_f = 2
n_f = 3
n_f = 4
n_f = 2
n_f = 3
n_f = 4
n_i = 2
0.456889
1.254407
-0.45367
0.004949
0.515368
-0.59421
-0.27262
0.082115
n_i = 3
-0.61239
0.162832
-0.7182
-0.48164
-0.21837
-0.79069
-0.62482
-0.44184
n_i = 4
-0.74951
-0.51437
-0.81789
-0.66502
-0.49488
-0.86474
-0.75754
-0.63929
n_i = 2
0.033628
1.003917
2.100884
0.382283
1.084354
-0.44184
0.00049
0.488424
n_i = 3
-0.43808
0.089416
0.685775
-0.59147
0.133145
-0.69656
-0.45609
-0.19083
n_i = 4
-0.62735
-0.27753
0.117957
-0.72907
-0.50165
-0.79877
-0.63929
-0.46338
n_i = 2
0.391609
1.697943
3.174828
0.011729
0.861016
1.806238
0.346994
1.003917
n_i = 3
-0.22364
0.505144
1.329077
-0.43557
0.038234
0.56556
-0.58077
0.117957
n_i = 4
-0.47815
0.011729
0.56556
-0.6206
-0.30212
0.052339
-0.7182
-0.49488
89
In Table 4.9, there were a few values close to 0, which suggested the resulting
data rate recovery outperforms the other three modulation schemes. Similarly to
the previous discussions, positive values indicated the adaptive factor was greater
than the interference factor, where negative values indicated the reverse situation.
The absolute value of the subtraction can determine the efficiency of the adaptive
factors. It measured how closely the adaptive factors can approach the interference
factors. The adaptive efficiency can be demonstrated in the following Figure 4.8
below:-
3.5
Initial ni = 2, Mi = 2
Initial ni = 3, Mi = 2
Initial ni = 4, Mi = 2
Initial ni = 2, Mi = 3
2.5
Initial ni = 3, Mi = 3
Initial ni = 4, Mi = 3
Initial ni = 2, Mi = 4
Initial ni = 3, Mi = 4
Initial ni = 4, Mi = 4
1.5
0.5
-0.5
-1
4
5
6
7
Final nf Values According to M f = 2, 3, 4
90
RM n PAPM _ f RM n PAPM _ i
( M i 1) 2
2
1
ni log 2 ( M i ni ) Q ( BERM n PAPM _ i )
(4.17)
Q 1 ( BER
( M f 1) 2
M n PAPM _ f )
n f log 2 ( M f n f )
For severe degradation, the above operation was again performed with BERM-n-8
PAPM_i=10
process will not be repeated, and keeping the remaining parameters unchanged,
final data rate values can be obtained:
Final Rb (Mbps)
3.9
249
Initial Rb (Mbps)
4
250
Final Rb (Mbps)
3.9
248
From the results in the table, referred to Table 4.1, Table 4.4 and Table 4.7 for
OOK, L-PAM and L-PPM modulation schemes, the efficiency of data rate
recovery for the adaptive M-n-PAPM scheme can be further demonstrated, as in
above Table 4.10.
91
4.3.1
92
Figure 4.9 OOK and L-PAM SNR vs BER comparison (with L=2, 3, 4, 5)
In Figure 4.9, under same noise conditions, the PAM modulation scheme required
more optical power to achieve same level of BER compared with OOK. This
suggested that the PAM modulation scheme was not preferred for power limited
applications, e.g. mobile device and PDAs, where battery life was essential for
operation. However, PAM can be found useful in applications where bandwidth
was limited but not the power consumption. Since the PAM modulation scheme
required less bandwidth than OOK and PPM (e.g. when > 2 ), which
suggested that the PAM can provide a higher throughput.
93
In this Figure 4.10, the BER increased with ceiling height H, which suggested the
impact of ISI on the channel depending on room geometry. Although the H value
was larger than for the typical office set up, e.g. room height usually fell in the
range of 2-3 metres, this indicated how the BER performance can be affected by
the room geometry change at a low data rate, e.g. =1Mb/s. Detailed procedures
and Matlab program for Figure 4.10 can be found in Appendix IV-6. At higher
data rate, the significance of the contribution from H increased accordingly. This
can be illustrated in the following figure.
94
In Figure 4.11, the BER increased steadily when the data rate was higher than
about 21Mb/s. This suggested at lower data rate (less than the data rate threshold
which started increasing the BER), for a given room geometry and SNR value, the
OOK and 2-PAM modulation scheme can maintain a required BER with data rate
up to the threshold. When the data rate exceeded the threshold, the impact of
the ISI cannot be neglected. Detailed procedures and Matlab program for
obtaining Figure 4.11 can be found in Appendix IV-7.
95
The threshold data rate can also be obtained by an analysis method. When an
OOK or 2-PAM modulation scheme was employed, each pulse train contained
one information bit (1 or 0). Since the room height =3.5 (m), the shortest
time for an optical pulse travelling to and from the ceiling was = 2/, where
is speed of light, the highest sampling frequency without ISI interference was
then = 1/ = /2, thus the maximum available bandwidth was = /2 =
/4 =21.4 (MHz). When comparing combined power and bandwidth
requirement for different modulation schemes, the normalized bandwidth
requirement for both OOK and 2-PAM were 1, that was, = , so the
maximum achievable data rate at a given BER was 21.4 (Mb/s), which matched
well with the simulation results.
4.3.2
Consider the L-PPM modulation schemes under multipath ISI and short noise
only. Using equation (3.35), the BER performance of a 2-PPM modulation system
under different data rate can be simulated, similar to OOK and 2-PAM case,
H=3.5m. As discussed early, the L-PPM can be treated as special case of M-nPAPM where M=1. Thus the 2-PPM scheme can be treated as 1-2-PAPM
modulation schemes.
96
In Figure 4.12, compared with the OOK and 2-PAM, the 2-PPM can only
maintain half of the reliable data rate for a given BER when other channel
parameters were same. This was because the 2-PPM required twice the bandwidth
of the OOK and 2-PAM according to the work detailed in Chapter 3. The L-PPM
modulation scheme required less average power to maintain the same BER level
as the OOK and L-PAM (e.g. > 2 ), which came at a cost of higher
bandwidth requirements, thus resulting in a lower data rate. The threshold data
rate analysis for 2-PAM still held for the 2-PPM, since the 2-PPM required twice
the bandwidth as 2-PAM, so the threshold data rate for 2-PPM was half that of the
2-PAM. = 21.4/2 = 10.7/. The next Figure 4.13 was a zoomed version
of the previous Figure 4.12, and this again matched well with the simulation
results. Detailed procedures and Matlab programs for Figure 4.12 and Figure 4.13
can be found in Appendix IV-8 and Appendix IV-9 respectively.
97
98
Conclusions
In this chapter, the proposed adaptive modulation concepts were validated in the
context of stabilising the transmission data rate. This was a new attempt for
reducing the interferences presented to the channel by actively updating the
modulation orders. In fact, most modulation schemes proposed for the optical
wireless channel can be validated as candidates for adaptive modulations. Results
presented here can be used to further demonstrate the capability of adaptive
modulation using different signal modulation techniques. The analytical model
and simulation results helped confirming the feasibility of the adaptive modulation
techniques which can be used for the optical wireless channel.
99
Chapter 5
5.1
Introduction
5.2
System Structure
5.3
5.4
5.5
5.1
5.2.1
Fuzzy Sets
5.2.2
Membership Function
5.2.3
5.2.4
Fuzzy Rules
Model Parameters
5.3.2
5.3.3
ANFIS Model
5.4.1
System Structure
5.4.2
5.4.3
5.4.4
5.4.5
5.4.6
Results Comparison
Introduction
Artificial intelligence (AI) attracted much attention from both scientists and
engineers since it was discussed by John McCarthy in 1956 [98]. It received fast
100
growth by employing programs and algorithms that can imitate the activities of a
human brain such as reasoning, learning and pattern recognition [99].
In the AI hierarchy, there were categories including: logical AI, inference, genetic
programming, heuristics, pattern recognition and so forth [100]. Yet more general
classifications of AI applications can be divided into three main branches:
cognitive robotics, computational intelligence and data mining. These included the
sub categories which can be demonstrated in the Figure 5.1 that follows.
Artificial
Intelligence
Cognitive
Robotics
Neural
Networks
Computational
Intelligence
Fuzzy Logic
System
Evolutionary
Computation
Evolutionary
Algorithms
Data Mining
Artificial Immune
Systems
Concept
Mining
Text
Mining
Immunocomputing
Swarm
Intelligence
From the figure, among the AI family, cognitive robotics and data mining
emphasised on simulating human brain activity and the learning process
respectively [101, 102]. They were commonly used for solving complex problems
or assisting in the study of the learning process. Under the computational
intelligence category, neural networks produced a similar output in response to
their training or learning process [103, 104]. Evolutionary computation involved
101
an iterative search until a preset target was reached [105]. The artificial immune
system can often find applications in medical and biology analysis [106]. The
fuzzy logic (FL) control algorithm was a rule based approach which allowed
conditions to be expressed in natural language forms [107]. This enabled the
transfer of previous experience into automatic control processes. FL also exhibited
the simplest structure compared to other AI techniques, delivering a faster
response. Hence, FL was a good candidate technique that can be used to assist
adaptive modulation.
102
5.2
System Structure
The FL concept was theorised by Lotfi Zadeh in [108], yet the origins can be
traced back to ancient times with most applications of FL being control related
systems. A FL control system can be applied to adaptive modulation according to
Figure 5.2. In this section, the elements of the fuzzy system were discussed. The
following Figure 5.3 represented the basic structures of a fuzzy system:
Fuzzy Logic
System
Fuzzy Sets
Membership
Functions
Rules
Logical
Operations
103
5.2.1
Fuzzy Sets
In Figure 5.3, there were four elements in a fuzzy system. Compared to classical
sets, fuzzy sets (FS) were sets without a clearly defined boundary, which
contained elements with partial degrees of a membership [109]. FSs were
important part of FL systems. System states of different modulation schemes and
BER conditions of a specific channel can be grouped into different FSs. For
example, BER degradation can be grouped into two categories: moderate and
severe. The degree of degradation can also be grouped by its variation to the
required BER value, e.g.
5.2.2
Membership Function
104
the rate of level change according to BER variation, a sample value can be found
in Table 5.1 [71]:
5.2.3
No Change
0
Change Slow
10-50
Change Fast
>50
Complement: A ( x) 1 A ( x)
Intersection: A B ( x) min{ A ( x), B ( x)}
Union: A B ( x) max{ A ( x), B ( x)}
A and B . More detailed fuzzy set operators can be found in Appendix V-1.
105
5.2.4
Fuzzy Rules
Figure 5.4 Fuzzy logic system block diagram (figure adapted from [109])
In Figure 5.4, the FC system worked in sequential steps: 1.Convert input data to
FS (Fuzzify Inputs); 2. Apply fuzzy rules (FR) to FS (Fuzzy Logic Operation); 3.
106
Convert results to output data (Defuzzify Outputs). The fuzzy inference was the
process of formulating the mapping from a given input to an output using fuzzy
logic [107]. Detailed fuzzy operations were discussed in next section.
Model Parameters
For adaptive modulation, the modulation parameter change was based on the
discussion in Chapter 4 to find the best fit to maximise the data rate under
different BER variation caused by different noise source discussed in Chapter 2.
BER values can be grouped by: Minor, Moderate and Severe, which represent
changes to one, two and three orders of magnitude, respectively. The BER
variation can be normalised as:
BERunit log10 (
BERfinal
BERinitial
(5.1)
Where BERunit is the mapped fuzzy control input, BERinitial is the initial system
BER and BER final is the final system BER. By using equation (5.1), the new BER
unit was outlined in Table 5.2.
Table 5.2 BER degradation mapping
Modulation Level Change
Actual BER Change
(compared to original level)
BERfinal
log10 (
)
BERinitial
Minor
10
Moderate
100
Severe
1000
107
5.3.2
The rules defined range for the BER variation (input) and the required level
change (output). The BER variation range took the unit value given in Table 5.2.
The required level change referred to the value range given in Figure 4.8, e.g.
between [0, 5]. The membership functions of BER variation and level change can
be obtained according to its value range. By mapping BER variation and level
change values from inputs to degree of memberships, the input data were
converted to FS inputs as demonstrated in the previous Figure 5.4. This
relationship can be further demonstrated in Figure 5.5.
108
From Figure 5.5, the input BER variation value in the range [1, 3] was mapped
into FSs by its related three membership functions. The obtained degree of
membership (y axis) reflected which category the input BER variation belonged to.
For example, if the BER variation was 1, according to its membership function,
the resulting fuzzy set input was 1, which is the Minor case. Increasing the BER
variation value from 1 to 2, decreased the contribution from the Minor and
increased the Moderate contribution from 0 to 1. This was also true for the
Severe case, which suggested that depending on the input BER variation value
and its membership functions, the obtained fuzzy sets can represent input value
according to preset membership functions.
The output of the fuzzy control system was the level change requirement defined
in the group as zero, minor and large. The level change requirement value
109
can be obtained by mapping the input fuzzy sets using similar method as the BER
variation. The differences were the level changing range. This can be presented in
Figure 5.6. The resulting fuzzy system block diagram was shown in Figure 5.7.
110
111
From Figure 5.8, the relationship between BER variation and the required level
change was similar to a step function. The fuzzy system gave level change
instructions according to the channel status, e.g. the BER variations. Detailed
parameters of system A can be found in Appendix V-2.
5.3.3
In system B, Rate is how fast the BER changes. The first rules was same as in
system A, rule 2 to 5 was the joint impact between BER variation and the rate
change. The fuzzy inference process for system B was shown in Figure 5.9 [115].
112
In Figure 5.9, BER = 1.5 and Rate = 0.35 were chosen as sample inputs. The
inputs were first mapped to the FS using MF. The resulting FSs then followed the
fuzzy operator to get antecedent values for each rule. The consequent value of
each rule can be obtained by the implication operation (min operator). The output
FS can then be calculated by applying the aggregation operator (max operator) to
the consequent value. The required modulation level change can be obtained by
defuzzifying the aggregated FS using the centroid operation. The required level
change was 2.17.
113
state information, e.g. BER variation and its rate of changes. The block diagram
for the system B was outlined in Figure 5.10.
In Figure 5.10, system B included two inputs and one output, BER variation and
level change included three MFs while rate contained two MFs (numbered).
Both system A and system B were Mamdani type fuzzy inference model, where
the output MFs applied centroid calculation [116]. This was different from the
Takagi-Sugeno-Kang (TSK) model where the output MFs were either linear or
constant [117]. Comparing these two models, the TSK model was more
computational efficient and thus fit better in optimisation and adaptive control
systems. While the Mamdani model was efficient in capturing the expert
knowledge but was computational inefficient, thus can be used for a well known
114
system modelling where expert knowledge was available. The input/output for
system B is shown in Figure 5.11.
115
System Structure
= + +
(5.2)
where , were inputs, is system output and , , are constants. The output
was weighted by a rule weight (firing strength) of the rule. The final output was
the weighted average of all system outputs and took the form
where is the number of rules [115]. An example ANFIS rule operation can be
illustrated in the following Figure 5.12.
Figure 5.12 ANFIS rule operation example (figure adapted from [110])
In Figure 5.12, 1 and 1 were membership functions for input 1 and input
2, since the final output was averaged over all output values using weight . By
116
changing , the output of the system can be adjusted to better fit the model data
sets [120].
The ANFIS system can employ either back propagation gradient descent or
combined least-squares and the back propagation method (hybrid) to obtain the
FIS structure [115]. The benefits included a more sophisticated system structure
best fit to the data sets, no need to understand system behaviour prior modelling
(black box) and flexible system adjustment methods for updating parameters [119].
5.4.2
ANFIS was especially useful when the input and output data were available for a
fuzzy system, e.g. modulation adaptation instructions for a specific channel set up.
Without prior knowledge of the membership functions, the ANFIS model can
identify a good approximation to the fuzzy inference system by learning the
known data set of that unknown system [111]. The previously developed fuzzy
system B can be used to demonstrate the benefits of the ANFIS model.
Obtain training and checking data from fuzzy system B, the obtained data set
containing 100 input and output pairs. The detailed data sets can be found in
Appendices V-4 and V-5. The training data and checking data were identical to
the fuzzy inference system yet had different ranges to provide comparability.
5.4.3
The input data for training and checking the ANFIS system can be grouped into
two categories. Singleton and 2-D recursive matrix data sets, where singleton data
117
set was obtained by sampling the BER variation and the rate change value along
its range linearly. The resulting output data was calculated from line values in the
input space. The singleton data sets can be used to determine the rough shape of
the unknown system, e.g. the output along a specific direction within the valid
input data sets. The singleton data sets can miss some of the system structures as it
offered less coverage of the input space.
5.4.4
The 2-D Recursive data set was obtained by fixing values of one system input and
listing all values of other inputs. By repeating this process again to the next fixing
values of the chosen input, the 2-D recursive data set was obtained. The 2-D
recursive data set included all data values according to the chosen input data range.
It can provide a better approximation to the unknown system compared to
singleton data sets. The generation process of the 2-D recursive data set was more
complex than the singleton data sets. The comparison between singleton and 2-D
recursive process was shown in Figure 5.13.
Figure 5.13 Comparison of single ton and 2-D recursive data set
generation
118
(a)
(b)
(c)
(d)
(e)
(f)
Figure 5.14 Singleton (a) BER variation (b) Rate value (c) Output levels and
recursive (d) BER variation (e) Rate value (f) Output levels
In Figure 5.14, the recursive data set covered more area than the singleton set as in
(e) and (f). In this test case, one of the inputs (rate) was chosen as the recursive
variable. The approximation accuracy of the trained ANFIS system can be
improved if all inputs were chosen as the recursive variable. The complexity of
the data set generation was also increased by a factor of !, where is
total number of input variables and is the data set resolution number. When
119
the total input numbers were relatively small (e.g. less than 10), it can be efficient
to use ( 1) number of input variables as recursive data set [120].
5.4.6
Results Comparison
The ANFIS system can be trained using four methods: 1. Back propagation
gradient descent (BPGD) only; 2. BPGD and one pass of least squared estimate
(LSE); 3.BPGD and LSE; 4. Sequential LSE. According to Jang [120], the hybrid
training method (third method) can provide best performance compared to other
method in terms of calculation complexity and accuracy. In this section, the
hybrid training method was used for training the ANFIS system along with the
BPGD method for comparison. The system parameters used for training can be
found in Table 5.3.
Value
20
MF numbers
[3 3]
Method used
Hybrid
BPGD
Grid partition
Data partition
Data resolution
100
Data structure
2 input 1 output
The training results can be found in following Figure 5.15 Figure 5.20.
120
121
122
123
Figure 5.15 and Figure 5.17 were for the ANFIS system trained using BPGD,
hybrid on a singleton data set. Figure 5.19 was for the trained ANFIS system by
employing the hybrid method on recursive data set. Figure 5.16, Figure 5.18 and
Figure 5.20 illustrated the respective training errors. Comparing Figure 5.15 with
Figure 5.17, it can be seen that the hybrid method captured more system
information than the BPGD method. However, the training errors of both methods
were high since they were limited by the singleton data set as reflected in Figure
5.16 and Figure 5.18. Comparing Figure 5.19 with Figure 5.11, the recursive data
set captured most system structure information from the training data, and was
identical to the system it approximated. According to Figure 5.20, the training
error of the recursive data set was negligible compared with ANFIS system
obtained using singleton data set.
Employing the neuro-fuzzy approach via ANFIS was also demonstrated. This can
be used to model unknown system model when given sufficient input/output data.
By adjusting system parameters using different learning algorithms, ANFIS can
provide excellent approximation to the unknown system. This was particularly
124
Conclusions
The artificial intelligence control methodologies were combined with adaptive
modulation schemes. Fuzzy logic control was selected as a viable control process
that provided simple yet powerful control functionality. Simulation results
confirmed the characteristics of the obtained fuzzy-controlled-adaptivemodulation schemes. By equipping the newly developed adaptive modulation
schemes, communication systems can provide flexible yet efficient adaptations for
improving transmission throughput.
The new concept exploited some exciting features, such as the ability to balance
between power and bandwidth requirements, increased immunity to different
types of interferences by optimising modulation parameters, ability to update
control patterns through training. This certainly can attract more attentions from
the mobile device design engineers, and contribute to the developments on more
efficient modulation schemes for the optical wireless industry.
125
Chapter 6
6.1
Introduction
6.2
System Reliability
6.3
6.2.1
Variable ISI
6.2.2
6.2.3
6.1 Introduction
The reliability or robustness of a communication system usually referred to the
ability to maintain certain system performances under interferences [29]. As
discussed in Chapter 2, the optical wireless channel can be easily affected by
channel uncertainty. For example, distance between transmitter and receiver,
distance from ambient light source or optical propagation path changes can result
in BER variation. Using the adaptive modulation model and FL control concept
developed in Chapter 4 and Chapter 5, different system interference can be
addressed within the context of system reliability. In this chapter, the utility of the
adaptive modulation schemes that employ FL control were further demonstrated.
126
6.2
System Reliability
127
Figure 6.1 BER and data rate performance for M-n-PPM (M=1, n=4)
modulation scheme with variable H and no ambient light interference
From Figure 6.1, it can be seen that the variation of H can affect BER
performance at different data rate. Since the background ambient light noise was
not considered, the factors contributing to the BER variation were thus purely the
consequences of channel geometry variations. In contrast to other approaches to
combat the ISI variation, the adaptive modulation system in this work can change
its modulation parameters to provide compensations for H variation. In this case, n
can be updated to find a match. Considering M-n-PAPM (M=1, n=4) with a data
rate of 50.5 Mb/s as a benchmark, the adaptation search results was shown in
Table 6.1
128
2.910-6
1.810-4
3.010-12
3.410-8
1.8 10-5
46.9
46.9
46.9
0.6
3.1
3.1
BER
3.110-16
3.910-10
2.110-6
43.5
43.5
43.5
1.1
6.6
6.5
BER
2.510-3
1.010-2
2.910-2
50.5
50.5
50.5
0.08
0.1
0.5
BER
1.010-3
4.010-3
2.310-2
75.8
75.8
75.8
1.2
1.2
9.3
BER
Rb=50.5Mb/s, M=1, n=4
BER
M=1, n=5
M=1, n=6
M=1, n=2
M=2, n=4
In the table, various combinations of M and n values were listed and improved or
evenly matched system parameters were marked with light shading. Compared to
the benchmark, when H increased from 1m to 3m, BER can be maintained or
improved by increasing n thus increasing the pulse slot numbers within a symbol.
However, this came at a cost of reduced data rate, for instance it can be seen that
the data rate dropped to 46.9 Mb/s and 43.5 Mb/s for n = 5 and n = 6 respectively,
representing 7% and 14% data rate losses for these cases. As would be expected,
the reverse happened when the n value decreased as the system was trading
bandwidth as n increased to decrease BER. As an example, using an n value of 2
129
maintains the data rate with H variation but offered a BER that was much worse
than the original one, e.g. for M=1, n=2 and H=2m, the BER was only 1.010-2,
which was unlikely to be tolerated by any system requirements. It was interesting
to note that when the amplitude level was increased, the data rate can be improved
significantly, e.g. when M=2, n=4, data rate improved to 75.8Mb/s which was
nearly 50% improvement compared to the original 50.5Mb/s. This was due to the
fact that the amplitude modulation was more bandwidth efficient. The increased
data rate was a consequence of reduced bandwidth requirements. Yet this also
came at a cost of more power consumption and reduced achievable BER, which
offset the benefit from the data rate increase. Moreover, the increased noise
susceptibility was absent from this scenario as the ambient noise was ignored, the
room was dark. It can be observed that the simulation time was proportional to
the H and M values.
From Table 6.1 and the discussion above, when variable ISI was the main source
of system degradation, the adaptive modulation system can reduce the consequent
BER variation by increasing number of pulse positions within a symbol sequence.
This came at a cost of data rate loss and a trade off was necessary between loss of
data rate and BER.
130
section, for constant ISI (H=1), the effect of ambient light noise can be simulated,
with the results illustrated in Figure 6.2. It can be observed that when the ASR
value increased, the lower data rates were most affected. For example, a system
operating at a data rate of 10.5 Mb/s can only achieve a BER of 3.1 10-2 when n
was increased to 10. This was a severe degradation compared to the case when n =
1. The ambient light interference had a much smaller impact on the higher data
rates in comparison to the lower data rates. The simulation time of different ASR
values were comparable so this was different from the case with variable H, which
suggested simulation time was identical with H values.
Figure 6.2 BER and data rate performance for M-n-PPM (M=1, n=4)
modulation scheme with variable ASR and constant ISI (H=1m)
In Figure 6.2, when the ASR increased, the BER of system operating at lower data
rates increased significantly. Although higher data rate were also affected by high
131
ASR, the degradation were less severe compared to the lower data rates. It can
also be noticed that for a given ASR, there was a data rate value that can minimise
achievable BER which can be denoted by _ . When the ASR increased,
_ increased accordingly; here the resulting _ values were
20.5Mb/s, 50.5Mb/s and 120.5Mb/s for ASR values of 1, 10 and 50 respectively.
Taking a data rate of 50.5 Mb/s as a benchmark again, the adaptive M-n-PAPM
parameters can be updated to find an optimum combination to reduce the BER.
Similar to the variable ISI case, the pulse position numbers and amplitude levels
can be adjusted and detailed simulation results can be found in Table 6.2
132
2.910-6
1.810-4
3.310-12
3.510-10
1.9 10-2
46.9
46.9
46.9
1.2
1.3
1.2
BER
3.110-16
2.010-13
1.210-2
43.5
43.5
43.5
2.3
2.3
2.3
BER
2.610-3
3.510-3
4.410-2
50.5
50.5
50.5
0.1
0.1
0.1
BER
6.910-3
8.310-2
1.810-2
75.8
75.8
75.8
2.4
1.9
1.2
BER
Rb=50.5Mb/s, M=1, n=4
BER
M=1, n=5
M=1, n=6
M=1, n=2
M=2, n=4
50
This table illustrated similar results to those shown in Table 6.1, suggesting that
the interference induced by the ambient light noise can be reduced by changing
pulse position values. System improvements were marked with light shading.
Compared to the variable ISI case, the simulation time was shorter and
comparable between different ASR values except for the M = 2 and n = 4
combinations. By actively choosing system parameters, the adaptive modulation
can maintain the desired communication link requirements under variation of the
background ambient noise.
133
Comparing results obtained in Table 6.1 and Table 6.2, by changing the pulse
position values, the bandwidth requirements and thus the achievable system data
rate can be adjusted. When increasing the pulse position orders, the occupied
period of the on chip within a pulse sequence was reduced, thus the periodic
ambient light noise contribution can be reduced compared to lower order pulse
position modulation systems. Higher order pulse position systems can lead to
lower power consumption and better immunity for both ISI and ambient light
noise interference. The combined contribution from both pulse position and
amplitude was thus a balance among different system parameters. This can be
further investigated in the following section.
134
Table 6.3 Initial system parameters for adaptive M-n-PAPM (M=1, n=4)
modulation with H=1 and ASR=50
Parameters
4-PPM
(M=1, n=4)
Value
Minimum BER
7.4 10-4
_ (Mb/s)
120.5
27.9
In order to find the optimum combination of pulse position and amplitude level,
exhaustive search can be used with a limit on the maximum amplitude level of 2
and pulse position values allowed within [2, 16]. Four candidate search results can
be found in Table 6.4
Table 6.4 System parameters for adaptive M-n-PAPM (M=1, n=4)
modulation with H=1 and ASR=50 using exhaustive search
Parameters
4-PPM
(M=1, n=4)
2-4-PAPM
(M=2, n=4)
2-8-PAPM
(M=2, n=8)
2-9-PAPM
(M=2, n=9)
Value
Minimum BER
7.4 10-4
_ (Mb/s)
120.5
27.9
Minimum BER
2.210-4
_ (Mb/s)
180.8
56.2
Minimum BER
1.310-8
_ (Mb/s)
100.5
298
Minimum BER
7.310-10
_ (Mb/s)
93.1
217.5
135
From this table, when H=1, ASR=50, the minimum BER achieved for a 4-PPM
(M=1, n=4) modulation scheme was 7.4 10-4 with a data rate of 120.5Mb/s. The
BER was not acceptable when compared to the 10-9 requirements. However, by
increasing the pulse position number, 2-8-PAPM and 2-9-PAPM achieved
improved BER values of 1.3 10-8 and 7.3 10-10 with a reduced data rate of
100.5Mb/s and 93.1Mb/s, respectively. The 2-4-PAPM scheme achieved a higher
data rate of 180.8Mb/s but the BER performance was not as good as the other two
candidates. The BER performance of all data rate steps was shown in Figure 6.3.
10
4-PPM
2-8-PAPM
2-9-PAPM
-1
10
-2
10
-3
10
-4
BER
10
-5
10
-6
10
-7
10
-8
10
-9
10
-10
10
50
100
150
Figure 6.3 BER and data rate performance for candidate adaptive M-n-PPM
modulation scheme with ASR=50 and H=1m
From above figure, for specific system degradation, the adaptive modulation
scheme can be optimised according to different system requirements (in this test
case, finding the optimum BER). The SNR to BER performance can be compared
136
to the case where no system adaptation was performed (4-PPM). This can be
found in Figure 6.4.
In Figure 6.4, the candidate modulation schemes reduced the optical SNR by at
least 3dB to achieve a BER of 10-9 compared to modulation without adaptation.
The 2-9-PAPM required 0.3dB more power compared to 2-8-PAPM. This was a
result of balancing between pulse positions and amplitude levels. In this way the
adaptive M-n-PAPM modulation schemes can utilise the benefits from both PAM
and PPM modulation schemes. Depending on system requirements, the adaptive
modulation can thus provide robustness under channel uncertainty with negligible
impact on system performance.
137
An FL control system can be set up for the adaptive modulation system discussed
in previous sections. This system was named C and its rules can be expressed as
follows:
Where H is ceiling height, ASR is the ambient light noise to signal ratio, M and n
were the resulting amplitude level and pulse position change values. The system
had two inputs (H and ASR) and two outputs (M and n). Detailed fuzzy system
construction was similar to that in Chapter 5 and can be found in Appendix VI-1,
the original fuzzy system mapping can be found in Figure 6.5
138
Figure 6.6 ANFIS trained using hybrid with recursive data set (system D)
139
In Figure 6.5, system C can be used to give instructions for a 4-PPM modulation
system under different interferences. Since the simulation time increased with
channel geometry and amplitude level, the simulation can be done prior to system
installation to reduce response time. This can be realised by simulating a wide
range of degradation scenarios and storing the adaptive control instructions on the
system memory chips. When the system was deployed in the designated working
environment, the optimum modulation parameters can be adapted according to the
preset instructions. For applications that required optimisation or robustness under
a specific interference, the adaptation instructions can be optimised particularly
for that requirement to better suit the designated environment. Detailed parameter
of system D and the training data can be found in Appendices VI-2 and VI-3
The FL control system can realise system adaptation with just three rules applied
to fuzzy interference process. This showed the simple yet powerful approach the
FL method can provide for system design. From Figure 6.6 and Figure 6.7, the
ANFIS model can identify the required control pattern by learning from training
data set of an unknown system. This was extremely useful when the control
patterns were complex and no prior knowledge of the system design was
available.
6.3
Summary
The performances of the adaptive modulation system were discussed in the
presence of combined effects of ISI and background ambient noises. From the
analysis, both factors must be taken into account when validating modulation
140
Conclusions
The fuzzy-logic-controlled-adaptive-modulation-schemes were validated under
different types of interferences, simulation results showed the new schemes were
effective for reducing the system degradation. Depending on the interference
patterns, significant improvements can be achieved by validating the possible
combinations of modulation parameters. Modulation optimisation models can be
developed for specific and general applications. Previous developed system
control patterns can be used as references for new channel environment. The
capability of realising reliable communication through the fuzzy logic controlled
adaptive modulation can be observed.
141
Chapter 7
7.1
Conclusions
7.2
Future Work
7.1 Conclusions
This chapter concluded the thesis with main findings from the discussions in
previous chapters. The main task of this thesis was to identify areas where optical
wireless communication can be better employed by the adaptive modulation
techniques discussed in this thesis.
As stated in Chapter 1, the optical wireless channel can offer attractive benefits
over the RF channel. Optical wireless communication became an important
complement for the communication systems. The system architecture grew more
complicated and sophisticated. This resulted more demand for channel throughput
and robustness.
142
In Chapter 2, the unique channel model for the optical wireless communication
was discussed. Channel topologies and propagation model were demonstrated.
Most important, the interference artificial light model was described.
In Chapter 3, candidate modulation schemes for the optical wireless channel were
discussed. Comparisons between modulation schemes were carried out.
In Chapter 5, the fuzzy logic control method was applied to the adaptive
modulation scheme. The control model used in the test case showed the capability
of the fuzzy logic control process. The ANFIS model developed was validated
using different data set from the system, and provided excellent approximation to
model the unknown control pattern though training.
The main arguments of this thesis were based on the modulation optimisation. The
following contributions were made:
143
2. The adaptive modulation scheme was proposed and validated according to the
requirements of optical wireless communication systems. Comparisons of the
adaptive scheme with other schemes showed that the proposed system can better
exploit the throughput capacity under certain system degradations.
3. Fuzzy Logic control modules were developed for the adaptive modulation
scheme. The system can achieve self adaptation by using fuzzy inference methods,
which benefited in a simple system structure compared to other artificial
intelligence systems. The ANFIS model was very efficient in pattern identification.
1. The ultimate channel capacity of the optical wireless channel remains an open
question. Although the channel capacity has been extensively discussed in the
literature, general expressions for the ultimate optical channel capacity still
remain open. Further investigation on this topic will help the researchers to
better understand the optical wireless channel.
144
3. The fuzzy logic control algorism developed in this thesis can be further
investigated. As the efficient control mechanism will provide the
communication system with more accurate instructions for adapting system
parameters. Applications on different system requirements can also be
analysed using this model.
145
APPENDIX
Appendix II-1 Parameters and Geometry for Simulation (Unblocked)
Parameter
Room Dimensions (metre)
(length width height)
Coordination System Geometry
North (reflectivity) %
South
West
East
Ceiling
Floor
Coordinate (0 0 0)
Transmitter Location (metre)
Transmitter Elevation and Azimuth (degree)
Transmitter Lobe Order
Receiver Location (metre)
Receiver Elevation and Azimuth (degree)
Receiver Werea (2 )
Receiver FOV (degree)
Transmitted Optical Power (Watts)
Time Step (ns)
Resolution (K=1 bounces)
Resolution (K=2 bounces)
Resolution (K=3 bounces)
Value
532
X (North to South)
Y (East to West)
Z (Floor Ceiling)
0.58
0.56
0.12
0.30
0.69
0.09
South East
Floor Corner
(3 0.8 2)
(-90 0)
1
(2 2 0.8)
(60 0)
1
70
1
0.2
30
8
4
146
Parameter
Room Dimensions (metre)
(length width height)
Coordination System Geometry
North (reflectivity) %
South
West
East
Ceiling
Floor
Coordinate (0 0 0)
Transmitter Location (metre)
Transmitter Elevation and Azimuth (degree)
Transmitter Lobe Order
Receiver Location (metre)
Receiver Elevation and Azimuth (degree)
Receiver Werea (2 )
Receiver FOV (degree)
Transmitted Optical Power (Watts)
Time Step (ns)
Resolution (K=1 bounces)
Resolution (K=2 bounces)
Resolution (K=3 bounces)
Separator Corner Point (X Y Z)
Separator Dimensions
(length width height)
Separator Reflectivity
(north south; west east; ceiling floor)
Value
532
X (North to South)
Y (East to West)
Z (Floor Ceiling)
0.58
0.56
0.12
0.30
0.69
0.7
South East
Floor Corner
(3 0.8 2)
(-90 0)
1
(2 2 0.8)
(60 0)
1
70
1
0.2
30
8
4
(2.5 0 0)
(0.2 1 1.9)
(0.8 0.8; 0 0; 0 0)
147
1
2
= 1
1
2
1 1
=
2
2
2 2
2
2
2
2
2
2
1 1
=
2
() =
2
2
2
()
()
2 (1)
148
1
2
Since = 1 ()
= 1
Thus
= 1
Thus
1
2
=1
149
Appendix IV-1 Matlab program for calculating adaptive factor for PAM,
PPM and M-n-PAPM
In this program, there were four parts, which counted for OOK, L-PAM, L-PPM,
and M-n-PAPM schemes respectively. This can be shown in the following:
%OOK
%Moderate
sook=(qfuncinv(1e-7)/qfuncinv(1e-9))^2; %Calsulate OOK noise ratio
Rook=4; %data rate
Rookf=Rook/sook %data rate for new BER
pause
Rook=250; % data rate
Rookf=Rook/sook % data rate for new BER
%Severe
sook=(qfuncinv(1e-4)/qfuncinv(1e-7))^2; %calculate OOK Noise Ratio
Rook=4; % data rate
Rookf=Rook/sook %data rate for new BER
pause
Rook=250; % data rate
Rookf=Rook/sook % data rate for new BER
%L_PAM
LPAM=[2:4]; %L-PAM modulation order
Ratio=(LPAM-1)./(sqrt(log2(LPAM))); % L-PAM adaptive factor
expression
%Calculate L-PAM adaptive ratio factor
for n=1:3,
for M=1:3,
Ratio2(n,M)=(Ratio(n)/Ratio(M))^2;
end
end
%Moderate
%Calculate L-PAM noise ratio
sm=(qfuncinv(1e-7)/qfuncinv(1e-9))^2;
%Calculate combined ratio factor
optimum(i,j)=Ratio2(i,j)*sm;
optimum =(round(optimum.*10))./10; %keeping one digit after decimal
optimum2=(optimum-1);% compare to the constant value 1
%convert diagonal value to 10, make it less identical to other values
ratio
150
for ii=1:3,
for jj=1:3,
if ii==jj
optimum2(ii,jj)=10;
end
end
end
find_minimum_value(abs(optimum2)) % find miminum adaptive compare ratio
factor and its location
%Severe
%repeat above process but with new sm value
sm=(qfuncinv(1e-4)/qfuncinv(1e-7))^2;
%data rate
%Moderate
Ri=250;
Rpam_m=Ri*Ratio2(2,3)*sm;
%Severe
ss=(qfuncinv(1e-4)/qfuncinv(1e-7))^2;
%data rate
Ri=250;
Rpam_s=Ri*Ratio2(2,3)*ss;
%LPPM
LPPM=[2:4];% Calculate L-PPM modulation order
Ratioppm=(LPPM.*log2(LPPM)); % Calculate L-PPM Ratio Factor
for n=1:3,
for M=1:3,
Ratioppm2(n,M)=(Ratioppm(n)/Ratioppm(M));
end
end
%Moderate
sm_ppm=(qfuncinv(1e-7)/qfuncinv(1e-9))^2; % Moderate BER Ratio
optimum_ppm=Ratioppm2*sm_ppm; % Moderate Data rate value matrix
optimum_ppm2 =(round(optimum_ppm.*100))./100; %keeping one digit after
decimal
optimum3_ppm=(optimum_ppm2-1);% compare to the constant value 1
%convert diagonal value to 10, make it less identical to other values
151
for ii=1:3,
for jj=1:3,
if ii==jj
optimum2_ppm(ii,jj)=10;
end
end
end
find_minimum_value(abs(optimum2_ppm))
compare ratio factor and its location
find
miminum
adaptive
%Severe
sm_ppm=(qfuncinv(1e-4)/qfuncinv(1e-7))^2;
%repeat same process but with sm_ppm new values
%data rate
Rppm_i=250;
Rppm_m=Rppm_i*Ratioppm2(2,3)*sm_ppm;
%Severe
ss=(qfuncinv(1e-4)/qfuncinv(1e-7))^2;
Rppm_i=250;
Rppm_s=Rppm_i*Ratioppm2(2,3)*sm_ppm;
%M-n-PPM
% Calculate M-n-PAPM ratio modulation matrix
M=2;
for i=1:3,
Ratiopapm(i)=(M+1)^2./((i+1).*log2(M.*(i+1)));
end
M=3;
for i=4:6,
Ratiopapm(i)=(M+1)^2./((i-2).*log2(M.*(i-2)));
152
end
M=4;
for i=7:9,
Ratiopapm(i)=(M+1)^2./((i-5).*log2(M.*(i-5)));
end
%Find adaptive ratio factor table 4.8 for M-n-PAPM
for j=1:9,
for k=1:9,
Ratio_papm_2(j,k)=Ratiopapm(j)./Ratiopapm(k);
end
end
%Find the optimum adaptive ratio factor
%First convert the diagonal value to 10, make it less significant
for ii=1:9,
for jj=1:9,
if ii==jj
Ratio_papm_2(ii,jj)=10;
end
end
end
%Moderate condition for M-n-PAPM
NRatio_mod_papm=(qfuncinv(1e-7)/qfuncinv(1e-9))^2;
% Calculate the data rate value matrix
optimum_papm =Ratio_papm_2*NRatio_mod_papm;
% Find the adaptive factor matrix
optimum_papm2 = optimum_papm -1;
%Convert diagonal back to 1 for table presentation
optimum_ppm2 =(round(optimum_ppm.*100))./100; %keeping one digit after
decimal
optimum3_ppm=(optimum_ppm2-1);% compare to the constant value 1
optimum_papm_Fig=optimum_papm2;
for i1=1:9,
for j1=1:9,
if i1==j1
optimum_papm_Fig(i1,j1)=0;
end
end
end
optimum_papm3=abs(optimum_papm2);
%Find the minimum Value
papm_min=min(min(optimum_papm3));
%Find the initial and final levels
for i2=1:9,
for j2=1:9,
if optimum_papm(i2,j2)==min(min(optimum_papm3))
n_i=i2;
n_f=j2;
end
end
end
%Severe
Rppm_i=250;
Rppm_s=Rppm_i*Ratioppm2(2,3)*sm_papm;
154
Appendix IV-2 Procedures and Matlab program for obtaining Figure 4.3
= =
2
(3.5)
(3.9)
In order to get normalised power and bandwidth for L-PAM, first the normalized
power and bandwidth equation can be obtained using above equation (3.5) and
equation (3.9), divide equation (3.5) with bandwidth requirement of OOK
modulation scheme and divide equation (3.9) with power requirement of OOK
modulation scheme the expression for normalised bandwidth and power
requirements for L-PAM scheme can be obtained as following:
= =
2
(IV-2 a)
1
2
(IV-2 b)
= 1010 (
)= 510 (
2
1 2
(1)2
)
=5
(
)
=5
(
)
10
10
2
2
(IV-2 c)
Where L is the amplitude levels of L-PAM scheme, equation (IV-2 a) is the x-axis
and equation (IV-2 c) is the y-axis. Take the L value from 2 to 16 with step 1, the
combined normalised power and bandwidth can be obtained. The Matlab program
thus is the following:
155
%L-PAM
L=[2:16];
x=1./log2(L);%bandwidth normalised to OOK
y=5*log10(((L-1).*(L-1))./(log2(L)));%power normalised to OOK
plot(x,y,'o:');
grid on
Increase x-axis range to 1.2, and add legends, Figure 4.3 can be obtained
156
(3.10)
(3.11)
(IV-3 a)
(IV-3 b)
= 1010
=510
= 510
= 510
2
2
=510
= 510 0.52
(IV-3 c)
Where L is the slots number of L-PPM scheme, equation (IV-3 c) is the y-axis and
equation (IV-3 b) is the x-axis. Increase L value from 2 to 16, the combined
normalised power and bandwidth can be obtained. The Matlab program thus is the
following:
%L-PPM
L=[2:16];
x=L./log2(L);
y=-5*log10(0.5*L.*log2(L));
plot(x,y,'s:');
grid on
157
Add legends and labels to the above figure, Figure 4.5 can then be obtained
158
2 2
2
(3.12)
(3.13)
2 2
(IV-4 a)
(IV-4 b)
1010 (
)= 510 (
2
) =510 (
2 2
2
2 2
)2 =510 (
(IV-4 c)
Where M and n is the amplitude level and slots number of M-n-PAPM scheme
respectively, equation (IV-4 c) is the y-axis and equation (IV-4 b) is the x-axis.
Increase M and n value from 2 to 16, this is done by fix one of M or n value first
and increase the other unfixed variable from 2 to 16, repeat this process, then the
combined normalised power and bandwidth can be obtained. Together with the
normalised power and bandwidth requirements of L-PAM, L-PPM and OOK for
references, the Matlab program can be found as the following:
159
%L-PPM
x=L./log2(L);
y=-5*log10(0.5*L.*log2(L));
%OOK
%the reference of OOK is the point (1.0) as the OOK bandwidth and power
%requirement is normalised by other scheme
plot(xx,yy,'o:',x,y,'s:',1,0,'v'); %plot the normalised power and bandwidth
%requirements of L-PAM, L-PPM and OOK
%for reference
hold on
%M-n-PAPM
for M = 2:16 %fix M value
for n = 2:16 %repeat calculation for the entire range of n
xpp=n./log2(n*M);
ypp=5*log10(2*M.*M./(n*log2(n*M)));
plot(xpp,ypp,'k.:');
grid on
end
end
hold off
160
Add legends and labels to the above figure, Figure 4.7 can then be obtained
161
1
=
1
2
1 1
0 =0
0 + 1
0 =1
(3.23)
where
1
0 =
1 =
,
=0
1
=
= = /
=1
1
+1
0 =0
0<<
+ 1
0<<
+ 1 ] (3.26)
0 =
1
=0
is the convolved
signal after the optical wireless channel. The detection threshold make hard
decision on received pulses. For a received pulse sequence, there were three types
162
2. Amplitude levels
Since 2-PAM has two possible levels of amplitudes and this is in line with the
definition
of
OOK.
By
changing
the
variable
Amax
in
The following functions were developed to simulate the BER, detailed programs
were listed after the function descriptions.
Function Descriptions:
1. main_PAM_SNR_BER
164
2. am_prepwere
3. am_vi
4. betaPortion
5. convolve
6. erfh
7. simulateThresholding
%=============================
% Prepwere parameters
%---------------------------166
a=2.0*ceilingHeight/300000000.0
avLength=L
aphabetCount=1;
for i=1:L
aphabetCount=aphabetCount*(Amax+1);
end
aphabetCount
M=log(aphabetCount)/log(2.0)
lambda=2.0/Amax
%Part II:
bitsPerChip=M/avLength
T=bitsPerChip/Bandwidth%chip duration
scaled_chip_length=T/a %
%------------------------------------------------------%estimation of size of sequence beta:
%Consider only significant remnants of impulse from the past
%and neglect small remnants from too distant past, it can be
%estimated rigidly based on preset accuracy
%- - - - - - - - - - - - - - - - - - - - - - - - - - - accuracyEps=1.0e-3 %preset accuracy 0.001 is a good choice,
%although can go further
hThresholdTs= (1.0/accuracyEps)^(1.0/6.0) - 1;
if hThresholdTs<1.0
hThresholdTs=1.0
end
hThresholdTs
%mark temporary variable with "w":
wtapsNumber = hThresholdTs/scaled_chip_length
tapsNumber = (floor(wtapsNumber)) + 1 %1 is taken for safety.
%tapsNumber sets a number of elements in arrary b, that is , a
%number of most distant chip from the past if to assign number
%0 to current symbol and count backward in time
%Convert SN from dB to numbers:
SN=exp( SNR/10.0*log(10.0))
%Adjust x-scale adopted in MatLab for erfc:
SN=SN/sqrt(2.0);
%=============================
beta=[1:tapsNumber];
bh=[1:tapsNumber+L];
%Create beta:
for k=1:tapsNumber
beta(k)=betaPortion(k-1, scaled_chip_length);
end
167
2. am_prepwere.m
%==========================================
% Ambient light from fluorescent light driven by electronic ballast
%==========================================
function am_prepwere()
global am_b
am_b=[1:20];
global am_c
am_c=[1:20];
global am_zeta
am_zeta=[1:20];
global am_fi
am_fi=[1:20];
global am_d %shifted from 0 to 1
am_d=[1:13];
global am_teta %shifted from 0 to 1
am_teta=[13];
global am_A1_reciprocal
global am_A2_reciprocal
%Fundamental frequency of high frequency component in (7), Hz:
global am_fh
global PI2
am_A1_reciprocal=1.0/5.9;
am_A2_reciprocal=1.0/2.1;
am_fh=37.5E3;
168
PI2=2*pi;
log10 = log(10);
for i=1:20
am_b(i)=exp( log10*( -13.1*log(100*i-50) +27.1 )/20
);
am_c(i)=exp( log10*( -20.8*log(100*i) + 92.4 )/20
);
end
%Table I:Amplitude and phase parameters for low-frequency
%components.
amAux=[
1, 4.65, 0.00, 11, 1.26, 6.00, 2, 2.86, 0.08, 12, 1.29, 6.17,
3, 5.43, 6.00, 13, 1.28, 5.69, 4, 3.90, 5.31, 14, 0.63, 5.37,
5, 2.00, 2.27, 15, 6.06, 4.00, 6, 5.98, 5.70, 16, 5.49, 3.69,
7, 2.38, 2.07, 17, 4.45, 1.86, 8, 4.35, 3.44, 18, 3.24, 1.38,
9, 5.87, 5.01, 19, 2.07, 5.91, 10, 0.70, 6.01, 20, 0.87, 4.88
];
for ii=1:10
for jj=1:2
j=jj-1;
pos=(ii-1)*6+j*3+1;
i=ii+10*j;
am_zeta(i)=amAux(pos+1);
am_fi(i)=amAux(pos+2);
end
end
%Check results matrix
for i=1:10
j=i+10;
end
%Wong et al Table II. Amplitude and phase parameters for high%frequency components.
amAux2=[
0, -22.22, 5.09, 6, -39.30, 3.55, 1, 0.00, 0.00, 7, -42.70, 4.15,
2, -11.50, 2.37, 8, -46.40, 1.64, 3, -30.00, 5.86, 9, -48.10, 4.51,
4, -33.90, 2.04, 10, -53.10, 3.55, 5, -35.30, 2.75, 11, -54.90, 1.78];
for ii=1:6
for jj=1:2
j=jj-1;
pos=(ii-1)*6+j*3+1;
i=ii+6*j;
am_d(i)=amAux2(pos+1);
am_teta(i)=amAux2(pos+2);
end
end
3. am_vi.m
%==========================================
% Artificial light interference model
%==========================================
%Calculate ambient light average energy contribution over time t2-t1 starting
169
end
4. betaPortion.m
%Returns werea under h-function over interval [k*step,k*step+step]:
%
function retv=betaPortion(k, step)
t=k*step;%start time1
power=t+1;
value1=1/((power)^6);%calculate impulse response of time1
t=t+step;%time increment
power=t+1;
retv=value1-1/((power)^6);%calculate integration impulse response of time2time1
end
5. convolve.m
%Calculates outputPulses = inputPulses * beta starting from element "start " in
%array outputPulses
%where * is a discrete convolution:
function retv=convolve(start, inputPulses, outputPulses)
global tapsNumber
global beta
nOut=size(outputPulses,2);
nIn=size(inputPulses,2);
for k=start:nOut
s=0.0;
for j=1:tapsNumber
tail=k-j+1;
if(tail<1 || tail>nIn)
break;
end
s=s+inputPulses(tail)*beta(j);
end
outputPulses(k)=s;
end
retv=outputPulses;
end
6. erfh.m
%Works in two modes:
%When erfScale<0, returns flipped horizontally Heaviside Function.
%Otherwise, flips horizontally and shrinks erf by erfScale times.
function retv=erfh(x,erfScale)
if(erfScale<0)
171
7. simulateThresholding.m
%Finds BER for given channel over all chip sequences, all noise events, and all
%ambient light events.
function retv=simulateThresholding()
tapsSimulationLimit=31;
global a
global shotNoisePresented
global amSAR
global amInteferenceSummationPoints
global amInterferencePeriodTi
global Amax
global OOK_threshold
global SN
global b
global bh
global S
global tapsNumber
global lambda
global scaled_chip_length
if tapsSimulationLimit<=tapsNumber
sprintf('Taps Number limit exceeded.')
return;
end
%tapsLimit:
tL=min(tapsSimulationLimit,tapsNumber);
unitEventsCount=1;
eventsCount=1;
for i=1:tL
unitEventsCount=2*unitEventsCount;
eventsCount=eventsCount*(Amax+1);
end
unitEventsCount
eventsCount
172
EVENTS_MEASURE_LIMIT=1000000;
if eventsCount>EVENTS_MEASURE_LIMIT
message='Stat. events limit exceeded'
EVENTS_MEASURE_LIMIT
eventsCount
return;
end
teta=lambda*OOK_threshold%teta=lambda/2 for OOK lambda=1/2 for PAM,
%lambda=2/A
amK=1.0/amSAR % define ambient light to signal ratio amK
amInterferenceStep=amInterferencePeriodTi/amInteferenceSummationPoints;
BAmb=0.0;%set initial value
for iXAm=0:amInteferenceSummationPoints-1
t=amInterferenceStep*iXAm;
B=0.0; %initial BER value
Bup=0.0;
Bdown=0.0;
k=tL-1; %Probe slot to count statistics.
for e=0:unitEventsCount-1
%------------------------------%Generate signal of units:
mask=e;
weight=1;
for slot=0:tL-1
ampl=rem(mask,2); %ampl=bitand(uint32(mask),uint32(1));
%http://www.mathworks.com/access/helpdesk/help/techdoc/index.html?/access/h
elpdesk/help/techdoc/ref/bitshift.html&http://www.mathworks.com/access/helpde
sk/help/techdoc/ref/bitand.html
if ampl>0
weight=weight*Amax;
end
b(k-slot+1)=ampl;
mask=mask-ampl;
mask=mask/2; %mask=bitshift(mask,-1)
%http://www.mathworks.com/access/helpdesk_r13/help/techdoc/ref/uint8.html
end
%------------------------------for iW=0:weight-1
%---------------------------------------------%Decompose iW and assign amplitudes to a signal of units:
weightS=iW;
for j=1:tL
i=tL-j;
if(b(i+1)>0)
reminder=rem(weightS,Amax);
173
b(i+1)=reminder+1;
weightS=(weightS-reminder)/Amax;
end
end
%---------------------------------------------bh=convolve(k+1,b,bh);
%bh(k+1)=bh(k+1)
S=lambda*bh(k+1);
Z=S; %no ambient light yet (default value)
if amInteferenceSummationPoints>1 %add ambient noise if
%parameter >1
Z=Z+amK*am_vi(t, scaled_chip_length*a);
end
bk=b(k+1);
%shortcut for b matrix
nUp=1.0*(lambda*bk+teta-Z); %noise Up
nDown=1.0*(Z-(lambda*bk-teta)); %noise Down
SNRf=SN; %SNR factor
if(~shotNoisePresented) %no shot noise case
SNRf=-1;
end
up=0.0;%initial value for up noise contribution
down=0.0;%initial value for down noise contribution
%We have three principal cases, bk=0, bk=A, bk in the middle.
if(0==bk)
up=erfh(nUp,SNRf);
elseif(Amax==bk)
down=erfh(nDown,SNRf);
else
down=erfh(nDown,SNRf);
up=erfh(nUp,SNRf);
end
Bup=Bup+up;%Total Bup noise contribution
B=B+up;%add up Bup noise contribution to total noise
Bdown=Bdown+down;%Total Bdown noise contribution
B=B+down;%add up Bdown noise contribution to total noise
end % iW
end %for e
B=B/eventsCount
Bup=Bup/eventsCount
Bdown=Bdown/eventsCount
BAmb=BAmb+B/amInteferenceSummationPoints;
end
% for(int iXAm=0; iXAm<amInteferenceSummationPoints; iXAm++){
BAmb
retv=BAmb;
end
174
1. Ceiling height H
In order to find the variation of BER caused by multipath ISI, the ceiling height
considered here was within the range [0,20] meters. Since the interference from
artificial light not considered here, the room is dark. The contributions from the
ceiling height change will not cause significant variation in BER. A for loop was
used to calculate each BER under different ceiling height H values.
175
main_PAM_SNR_BER
%Simulation for BER vs Ceiling height H
Tic %program start
for ceilingHeight=1:20
%=============================
% Sub Model Usage
%----------------------------global shotNoisePresented
shotNoisePresented=1
global amInteferenceSummationPoints;
%global ambientLightPresented
%ambientLightPresented=0
%Set this parameter to 1 to disable ambient light:
%To enable ambient light, set this parameter to number of points over
%which
%averaging via interference interval is desired:
%Accracy is proportional to this number:
amInteferenceSummationPoints=1;
%=============================
%=============================
% Default parameters
%----------------------------%global ceilingHeight
global Amax
global L
global Rb
global SNR
global amSAR
global amInterferencePeriodTi
Amax=1 %Number of non-zero amplitude levels.
L=1 %Maximum number of chips in symbol
Rb=1E6
SNR=7 %Signal To Noise Ratio, db
amInterferencePeriodTi=25.0e-6 %In seconds.
amSAR=5 %Signal to Ambient light Ratio. = amSAR = 1/K where K is
parameter from [Wong at all].
global OOK_threshold; %In units of minumum non-zero chip.
OOK_threshold=0.5
%=============================
%=============================
%Derivative parameters:
global a %ISI length parameter in chips. Parameter of h-function.
global SN %SNR not in dB form:
global T %Chip length, seconds.
global avLength %Average number of chips in symbol.
176
beta=[1:tapsNumber];
bh=[1:tapsNumber+L];
%Create beta:
for k=1:tapsNumber
beta(k)=betaPortion(k-1, scaled_chip_length);
end
beta
am_prepwere();
simulateThresholding();
b(ceilingHeight)=simulateThresholding();
end
plot(b,'bv-');
toc %program end
178
1. Data Rate
Rb within range [1, 300] Mbps, increase step 20Mbps.
2. Ceiling height
H=3.5m
3. Amplitude levels
Amax=1
4. SNR value
As demonstrated in Figure 4.9, OOK and 2-PAM need 7.5dB to achieve a BER of
108 in the experiment setup. Thus SNR=7.5 dB
%global ambientLightPresented
%ambientLightPresented=0
%Set this parameter to 1 to disable ambient light:
%To enable ambient light, set this parameter to number of points over which
%averaging via interference interval is desired:
%Accracy is proportional to this number:
amInteferenceSummationPoints=1;
%=============================
%=============================
% Default parameters
%----------------------------global Amax
global L
global SNR
global amSAR
global amInterferencePeriodTi
ceilingHeight=3.5 %Height of the room.
Amax=1 %Number of non-zero amplitude levels.
L=1 %Maximum number of chips in symbol
SNR=7.5 %Signal To Noise Ratio, db
amInterferencePeriodTi=25.0e-6 %In seconds.
amSAR=1 %Signal to Ambient light Ratio. = amSAR = 1/K where K is
%parameter from [Wong at all].
global OOK_threshold; %In units of minumum non-zero chip.
OOK_threshold=0.5
%=============================
%=============================
%Derivative parameters:
global a %ISI length parameter in chips. Parameter of h-function.
global SN %SNR not in dB form:
global T %Chip length, seconds.
global avLength %Average number of chips in symbol.
global aphabetCount %Number of symbols in alphabet
global M %Bits per symbol
global bitsPerChip
global scaled_chip_length %T/a
global tapsNumber %"Memory" of multipath channel.
global beta %Discretized h., Array
global bh %Convolution b*h, Array
global lambda %(min non-zero Intensity)/average Intensity:
%=============================
%=============================
% Prepwere parameters
180
%---------------------------a=2.0*ceilingHeight/3e8
avLength=L
aphabetCount=1;
for i=1:L
aphabetCount=aphabetCount*(Amax+1);
end
aphabetCount
M=log(aphabetCount)/log(2.0)
lambda=2.0/Amax
%Part II:
bitsPerChip=M/avLength
T=bitsPerChip/Rb%chip duration
scaled_chip_length=T/a
%------------------------------------------------------%estimation of size of sequence beta:
%Consider only significant remnants of impulse from the past
%and neglect small remnants from too distant past, it can be
%estimated rigidly based on preset accuracy
%- - - - - - - - - - - - - - - - - - - - - - - - - - - accuracyEps=1.0e-3 %preset accuracy 0.001 is a good choice,
%although can go further
hThresholdTs= (1.0/accuracyEps)^(1.0/6.0) - 1;
if hThresholdTs<1.0
hThresholdTs=1.0
end
hThresholdTs
%mark temporary variable with "w":
wtapsNumber = hThresholdTs/scaled_chip_length
tapsNumber = (floor(wtapsNumber)) + 1 %1 is taken for safety.
%------------------------------------------------------%Convert SN from dB to numbers:
SN=exp( SNR/10.0*log(10.0))
%Adjust x-scale adopted in MatLab for erfc:
SN=SN/sqrt(2.0);
%=============================
beta=[1:tapsNumber];
bh=[1:tapsNumber+L];
%Create beta:
for k=1:tapsNumber
beta(k)=betaPortion(k-1, scaled_chip_length);
end
181
182
1
= 1 =
(3.35)
(3.36)
The L-PPM modulation scheme can be treated as a special M-n-PAPM with M=1.
The L-PPM employed MAP detection and can be demonstrated in the following
figure:
183
From above figure, received symbol can be recognised when noise contribution N
was not significant, that is, the noise not leading the value of wrong peak j greater
than value of correct peak i. When the received signal power exceed amplitude
level G, symbol error occurs, and when received signal power less than level G,
success detection is G= .
In order to simulate the BER for L-PPM scheme, Matlab program was written and
details were listed in the next section. The functions for L-PPM were listed as
following:
1. main_PPM
2. am_prepwere
3. am_vi
4. betaPortion
5. convolve
6. erfh
7. simulatePPM
Function 2,3,4,5 and 6 were as same as the case for L-PAM, since these were
common functions. The simulation variables for Figure 4.12 listed as following:
1. Data rate
Data rate for this case in the range of 1Mbps to 300Mbps, step is 10Mbps.
2. Ceiling height
Ceiling height H=3.5m
3. Pulse position slots number
L=2
4. SNR
Similar to L-PAM case, when no artificial light interference was considered, the
2-PPM requires 6dB to achieve BER of 108 .
184
Function 1 and 7 were different from L-PAM, and were listed below:
1. Function main_PPM
%Calculate BER for PPM with ISI and ambient light.
Tic %simulation time start
%for ceilingHeight=1:20
Rbindex=1;
for Rb=1e6:10e6:300e6
%=============================
% Noise Model Selection
%-----------------------------
global shotNoisePresented
shotNoisePresented=1
global amInteferenceSummationPoints;
%global ambientLightPresented
%ambientLightPresented=0
% Ambient light noise
amInteferenceSummationPoints=1;
% amInteferenceSummationPoints.
%
If this parameter>1 then ambient noise is taken into account,
%
and this parameter is number of integration points.
%
Set this parameter to 1 to disable ambient light.
%
Integration accracy is proportional to this parameter.
%=============================
%=============================
% Default parameters
%----------------------------global Amax
global L
global Rb
global SNR
global amSAR
global amInterferencePeriodTi
ceilingHeight=3.5%Height of the room.
Amax=1 %Number of non-zero amplitude levels.
L=2 %Maximum number of chips in symbol
%Rb=1e6
%SNR=7
SNR=6 %Signal To Noise Ratio, db
amInterferencePeriodTi=25.0e-6 %In seconds.
amSAR=0.05 %Signal to Ambient light Ratio. amSAR = 1/K where K is
%parameter from [Wong et al].
%=============================
185
%=============================
%Derivative parameters:
global a %ISI length parameter in chips. Parameter of h-function.
global SN %SNR not in dB form:
global T %Chip length, seconds.
global avLength %Average number of chips in symbol.
global aphabetCount %Number of symbols in alphabet
global M %Bits per symbol
global bitsPerChip
global scaled_chip_length %T/a
global tapsNumber %"Memory" of multipath channel.
global beta %Discretized h., Array
global lambda %(min non-zero Intensity)/average Intensity:
%=============================
%=============================
% Prepwere parameters
%---------------------------a=2.0*ceilingHeight/300000000.0
avLength=L
aphabetCount=L
M=log(aphabetCount)/log(2.0)
lambda=L
%Part II:
bitsPerChip=M/avLength
T=bitsPerChip/Rb
scaled_chip_length=T/a
%=============================
beta=[1:tapsNumber];
%Create beta:
for k=1:tapsNumber
beta(k)=betaPortion(k-1, scaled_chip_length);
end
beta %beta value
am_prepwere();
simulatePPM();
b(Rbindex)=simulatePPM();
Rbindex=Rbindex+1;
end
Rb=1:10:300;
semilogy(Rb,b,'kd-');
hold on;
toc %simulation time end
7. Function simulatePPM
%Finds P - symbol error probability (symbol error rate)
%for given channel over all chip sequences, all noise events, and all ambient light
events.
%It can be observed that BER=P/M.
%P is denoted as BAmb.
function retv=simulatePPM()
global a
global shotNoisePresented
global amSAR
global amInteferenceSummationPoints
global amInterferencePeriodTi
global Amax
global L
global M
global SN
global T
global S
global tapsNumber
global lambda
187
global scaled_chip_length
%------------------------------------------------%Prevent errors:
%- - - - - - - - - - - - - - - - - - - - - - - - tapsSimulationLimit=10000;
if tapsSimulationLimit<=tapsNumber
message='Taps Number limit exceeded.'
return;
end
if L<2
message='Incorrect value: L<2.'
return;
end
%- - - - - - - - - - - - - - - - - - - - - - - - %Prevent errors:
%------------------------------------------------SNR2=SN/sqrt(2.0); %Jump is built up "with" two noise events.
%Adjust x-scale adopted in MatLab for erfc:
SNR2=SNR2/sqrt(2.0);
%First, find out number of preceding symbols:
sslots=0;
%take enough slots to cover ISI tapsNumber:
while sslots*L<tapsNumber
sslots=sslots+1;
end
%Recalculated taps number occupied by preceding symbols:
pastTaps=sslots*L;
%Recalculate tapsNumber including primary symbol:
symTapsNumber=pastTaps+L;
b=[1:symTapsNumber];
bh=[1:symTapsNumber];
%Ambient contribution to primary symbol chips:
V=[1:L];
%Find out number of all combination of preceding symbols:
symbol_events=1;
for i=1:sslots
symbol_events=symbol_events*L;
end
%Now, symbol_events=L^sslots
%set limit to 1E6, this can be changed to larger value
PPMSymbolSimulationLimit=1000000;
%Protect against long calculations:
188
if PPMSymbolSimulationLimit<=symbol_events
sprintf('Symbol Slots Limit exceeded.');
symbol_events
PPMSymbolSimulationLimit
return;
end
%artificial noise factor
amK=1.0/amSAR;
weight_ISI_NOISE=1.0/symbol_events/L/(L-1);
BAmb=0.0;
amInterferenceStep=amInterferencePeriodTi/amInteferenceSummationPoints;
for iXAm=0:amInteferenceSummationPoints-1
tt=amInterferenceStep*iXAm;
BB=0.0; %"BER under integration sign" by time.
%Prepwere ambient contributions to current symbol:
for i=0:L-1
V(i+1)=amK*am_vi(tt+T*i, T);
end
for e=0:symbol_events-1
%------------------------------------%Generate symbols and chip sequences.
%- - - - - - - - - - - - - - - - - - mask=symbol_events;
for slot=0:sslots-1
sym=rem(mask,L);
mask=mask-sym;
mask=mask/L;
for i=0:L-1
b(slot*L+i+1)=0;
if sym==i
b(slot*L+i+1)=Amax;
end
end
end
%- - - - - - - - - - - - - - - - - - %Generate symbols and chip sequences.
%------------------------------------%Cycle through primary chips:
for i=0:L-1
%Fill primary symbol's chips with zeros:
189
for k=0:L-1
b(sslots*L+k+1)=0;
end
%Make i-th primary chip non-zero:
b(sslots*L+i+1)=Amax;
190
191
Operator Expressions
Equality
A (u ) B (u ) , u U
Union
Intersection
Complement
A (u ) 1 A (u ) , u U
Normalization
NORM ( A) (u ) A (u ) / max( A (u )) , u U
Concentration
Dilation
Intensification
2( A (u ))2
2
1 2(1 A (u ))
INT ( A) (u )
for
0 A (u ) 0.5
for
0.5 A (u ) 1
Algebraic Product
Bounded Sum
Bounded Product
Drastic Product
A (u )
A B (u ) B (u )
0
for B (u ) 1
for A (u ) 1
for A (u ), B (u ) 1
192
Name
Type
Inputs/Outputs
NumInputMFs
NumOutputMFs
NumRules
AndMethod
OrMethod
ImpMethod
AggMethod
DefuzzMethod
InLabels
OutLabels
InRange
OutRange
InMFLabels
OutMFLabels
InMFTypes
OutMFTypes
InMFParams
OutMFParams
Rule Antecedent
Rule Consequent
Rule Weight
Rule Connection
AdaptivePAPM01
mamdani
[1 1]
3
3
3
min
max
min
max
centroid
BER
Levels
[1 3]
[0 5]
Minor
Morderate
Severe
zero
minor
large
gauss2mf
gbellmf
gauss2mf
gbellmf
gbellmf
gbellmf
[0.033 0.87 0.1934 1.315]
[0.46 3.28 1.993 0]
[0.185 2.659 0.168 3.413]
[1.25 2.5 -2.776e-017 0]
[1.25 2.5 2.5 0]
[1.25 2.5 5 0]
123
123
111
111
193
Name
Type
Inputs/Outputs
NumInputMFs
NumOutputMFs
NumRules
AndMethod
OrMethod
ImpMethod
AggMethod
DefuzzMethod
InLabels
OutLabels
InRange
OutRange
InMFLabels
OutMFLabels
InMFTypes
OutMFTypes
InMFParams
OutMFParams
Rule Antecedent
Rule Consequent
Rule Weight
Rule Connection
AdaptivePAPM02
mamdani
[2 1]
[3 2]
3
5
min
max
min
max
centroid
BER, rate
Levels
[1 3] [0 1]
[0 5]
Minor Morderate Severe fast slow
zero small large
gauss2mf
gbellmf
gauss2mf
gbellmf
gbellmf
gbellmf
[0.183 1.06 0.183 1.34]
[0.46 3.28 1.993 0]
[0.183 2.664 0.168 3.454]
[0.5152 3.13 0.998 0]
[0.406 2.5 0.114 0]
[1.25 2.5 -2.776e-017 0]
[1.25 2.5 2.5 0]
[1.25 2.5 5 0]
[1 0] [2 1] [2 2] [3 1] [3 1]
13233
11111
11212
194
BER
1.0000
1.0202
1.0404
1.0606
1.0808
1.1010
1.1212
1.1414
1.1616
1.1818
1.2020
1.2222
1.2424
1.2626
1.2828
1.3030
1.3232
1.3434
1.3636
1.3838
1.4040
1.4242
1.4444
1.4646
1.4848
1.5051
1.5253
1.5455
1.5657
1.5859
1.6061
1.6263
1.6465
1.6667
1.6869
1.7071
1.7273
1.7475
1.7677
1.7879
Checking Data
rate
0
0.0101
0.0202
0.0303
0.0404
0.0505
0.0606
0.0707
0.0808
0.0909
0.101
0.1111
0.1212
0.1313
0.1414
0.1515
0.1616
0.1717
0.1818
0.1919
0.202
0.2121
0.2222
0.2323
0.2424
0.2525
0.2626
0.2727
0.2828
0.2929
0.303
0.3131
0.3232
0.3333
0.3434
0.3535
0.3636
0.3737
0.3838
0.3939
Level
1.9906
1.9814
1.9767
1.9756
1.9756
1.9756
1.9756
1.9756
1.9756
1.9756
1.9756
1.9757
1.9759
1.9761
1.9764
1.9769
1.9774
1.9781
1.9808
1.9874
1.9989
2.0155
2.0373
2.0643
2.0964
2.133
2.1739
2.2186
2.2658
2.3129
2.3584
2.4015
2.4413
2.4778
2.5106
2.5394
2.5645
2.5864
2.6059
2.6238
No.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
BER
1.0000
1.0152
1.0303
1.0455
1.0606
1.0758
1.0909
1.1061
1.1212
1.1364
1.1515
1.1667
1.1818
1.1970
1.2121
1.2273
1.2424
1.2576
1.2727
1.2879
1.3030
1.3182
1.3333
1.3485
1.3636
1.3788
1.3939
1.4091
1.4242
1.4394
1.4545
1.4697
1.4848
1.5000
1.5152
1.5303
1.5455
1.5606
1.5758
1.5909
rate
0
0.0081
0.0162
0.0242
0.0323
0.0404
0.0485
0.0566
0.0646
0.0727
0.0808
0.0889
0.097
0.1051
0.1131
0.1212
0.1293
0.1374
0.1455
0.1535
0.1616
0.1697
0.1778
0.1859
0.1939
0.202
0.2101
0.2182
0.2263
0.2343
0.2424
0.2505
0.2586
0.2667
0.2747
0.2828
0.2909
0.299
0.3071
0.3152
Level
1.9906
1.9833
1.9786
1.9762
1.9756
1.9756
1.9756
1.9756
1.9756
1.9756
1.9756
1.9756
1.9756
1.9757
1.9757
1.9759
1.976
1.9763
1.9766
1.977
1.9774
1.978
1.9786
1.9795
1.982
1.9868
1.9943
2.0046
2.0179
2.0342
2.0535
2.0757
2.1008
2.1285
2.1588
2.1914
2.2261
2.2624
2.2991
2.3355
195
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
1.8081
1.8283
1.8485
1.8687
1.8889
1.9091
1.9293
1.9495
1.9697
1.9899
2.0101
2.0303
2.0505
2.0707
2.0909
2.1111
2.1313
2.1515
2.1717
2.1919
2.2121
2.2323
2.2525
2.2727
2.2929
2.3131
2.3333
2.3535
2.3737
2.3939
2.4141
2.4343
2.4545
2.4747
2.4949
2.5152
2.5354
2.5556
2.5758
2.5960
2.6162
2.6364
2.6566
2.6768
0.404
0.4141
0.4242
0.4343
0.4444
0.4545
0.4646
0.4747
0.4848
0.4949
0.5051
0.5152
0.5253
0.5354
0.5455
0.5556
0.5657
0.5758
0.5859
0.596
0.6061
0.6162
0.6263
0.6364
0.6465
0.6566
0.6667
0.6768
0.6869
0.697
0.7071
0.7172
0.7273
0.7374
0.7475
0.7576
0.7677
0.7778
0.7879
0.798
0.8081
0.8182
0.8283
0.8384
2.6404
2.6565
2.673
2.6903
2.7086
2.7276
2.7472
2.7673
2.7876
2.8077
2.8272
2.8461
2.8641
2.8812
2.8973
2.9123
2.9263
2.939
2.9508
2.9615
2.9713
2.9805
2.9892
2.9982
3.0077
3.0188
3.0321
3.0489
3.0702
3.097
3.1303
3.1709
3.2188
3.2755
3.34
3.4113
3.486
3.5625
3.6374
3.7095
3.7769
3.839
3.8952
3.9455
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
1.6061
1.6212
1.6364
1.6515
1.6667
1.6818
1.6970
1.7121
1.7273
1.7424
1.7576
1.7727
1.7879
1.8030
1.8182
1.8333
1.8485
1.8636
1.8788
1.8939
1.9091
1.9242
1.9394
1.9545
1.9697
1.9848
2.0000
2.0152
2.0303
2.0455
2.0606
2.0758
2.0909
2.1061
2.1212
2.1364
2.1515
2.1667
2.1818
2.1970
2.2121
2.2273
2.2424
2.2576
0.3232
0.3313
0.3394
0.3475
0.3556
0.3636
0.3717
0.3798
0.3879
0.396
0.404
0.4121
0.4202
0.4283
0.4364
0.4444
0.4525
0.4606
0.4687
0.4768
0.4848
0.4929
0.501
0.5091
0.5172
0.5253
0.5333
0.5414
0.5495
0.5576
0.5657
0.5737
0.5818
0.5899
0.598
0.6061
0.6141
0.6222
0.6303
0.6384
0.6465
0.6545
0.6626
0.6707
2.3709
2.4052
2.4378
2.469
2.4983
2.5258
2.5505
2.5732
2.5939
2.613
2.6308
2.6476
2.6636
2.679
2.6939
2.7087
2.7237
2.7393
2.7552
2.7714
2.7876
2.8037
2.8195
2.8348
2.8498
2.8641
2.8779
2.891
2.9035
2.9152
2.9263
2.9365
2.9461
2.9549
2.9631
2.9705
2.9773
2.9835
2.989
2.9942
2.9989
3.0034
3.0076
3.012
196
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
2.6970
2.7172
2.7374
2.7576
2.7778
2.7980
2.8182
2.8384
2.8586
2.8788
2.8990
2.9192
2.9394
2.9596
2.9798
3.0000
0.8485
0.8586
0.8687
0.8788
0.8889
0.899
0.9091
0.9192
0.9293
0.9394
0.9495
0.9596
0.9697
0.9798
0.9899
1
3.9902
4.0294
4.0537
4.0658
4.0774
4.0883
4.0987
4.1085
4.1178
4.1267
4.1349
4.1425
4.1497
4.1563
4.1625
4.1682
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
2.2727
2.2879
2.3030
2.3182
2.3333
2.3485
2.3636
2.3788
2.3939
2.4091
2.4242
2.4394
2.4545
2.4697
2.4848
2.5000
0.6788
0.6869
0.6949
0.703
0.7111
0.7192
0.7273
0.7354
0.7434
0.7515
0.7596
0.7677
0.7758
0.7838
0.7919
0.8
3.0166
3.0217
3.0277
3.0349
3.0437
3.0545
3.0679
3.0842
3.1039
3.1273
3.1548
3.1866
3.2228
3.2638
3.3097
3.3599
197
BER
1
1
1
1
1
1
1
1
1
1
1.2222
1.2222
1.2222
1.2222
1.2222
1.2222
1.2222
1.2222
1.2222
1.2222
1.4444
1.4444
1.4444
1.4444
1.4444
1.4444
1.4444
1.4444
1.4444
1.4444
1.6667
1.6667
1.6667
1.6667
1.6667
1.6667
1.6667
1.6667
1.6667
1.6667
Checking Data
rate
0
0.1111
0.2222
0.3333
0.4444
0.5556
0.6667
0.7778
0.8889
1
0
0.1111
0.2222
0.3333
0.4444
0.5556
0.6667
0.7778
0.8889
1
0
0.1111
0.2222
0.3333
0.4444
0.5556
0.6667
0.7778
0.8889
1
0
0.1111
0.2222
0.3333
0.4444
0.5556
0.6667
0.7778
0.8889
1
Level
1.9906
1.9908
1.9996
2.0406
2.1503
2.3575
2.4959
2.5245
2.5264
2.5264
1.9755
1.9757
1.9844
2.0251
2.133
2.3359
2.4709
2.4984
2.5
2.5
2.0283
2.0285
2.0373
2.0791
2.1934
2.4122
2.5572
2.5837
2.5852
2.5852
2.4176
2.4177
2.4273
2.4778
2.6326
2.8468
2.9456
2.963
2.9639
2.9639
No.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
BER
rate
1
1
1
1
1
1
1
1
1
1
1.1667
1.1667
1.1667
1.1667
1.1667
1.1667
1.1667
1.1667
1.1667
1.1667
1.3333
1.3333
1.3333
1.3333
1.3333
1.3333
1.3333
1.3333
1.3333
1.3333
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
0
0.0889
0.1778
0.2667
0.3556
0.4444
0.5333
0.6222
0.7111
0.8
0
0.0889
0.1778
0.2667
0.3556
0.4444
0.5333
0.6222
0.7111
0.8
0
0.0889
0.1778
0.2667
0.3556
0.4444
0.5333
0.6222
0.7111
0.8
0
0.0889
0.1778
0.2667
0.3556
0.4444
0.5333
0.6222
0.7111
0.8
Level
1.9906
1.9908
1.9938
2.0104
2.0553
2.1503
2.3152
2.4583
2.5149
2.5255
1.9755
1.9756
1.9786
1.9952
2.0397
2.133
2.2946
2.4343
2.4893
2.4993
1.9755
1.9756
1.9786
1.9952
2.0397
2.133
2.2946
2.4343
2.4893
2.4993
2.1079
2.108
2.1111
2.1285
2.1769
2.2854
2.4799
2.6365
2.6869
2.6953
198
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
1.8889
1.8889
1.8889
1.8889
1.8889
1.8889
1.8889
1.8889
1.8889
1.8889
2.1111
2.1111
2.1111
2.1111
2.1111
2.1111
2.1111
2.1111
2.1111
2.1111
2.3333
2.3333
2.3333
2.3333
2.3333
2.3333
2.3333
2.3333
2.3333
2.3333
2.5556
2.5556
2.5556
2.5556
2.5556
2.5556
2.5556
2.5556
2.5556
2.5556
2.7778
2.7778
2.7778
2.7778
0
0.1111
0.2222
0.3333
0.4444
0.5556
0.6667
0.7778
0.8889
1
0
0.1111
0.2222
0.3333
0.4444
0.5556
0.6667
0.7778
0.8889
1
0
0.1111
0.2222
0.3333
0.4444
0.5556
0.6667
0.7778
0.8889
1
0
0.1111
0.2222
0.3333
0.4444
0.5556
0.6667
0.7778
0.8889
1
0
0.1111
0.2222
0.3333
2.5
2.5001
2.5098
2.5607
2.7086
2.9123
3.0068
3.0234
3.0244
3.0244
2.5
2.5001
2.5098
2.5607
2.7086
2.9123
3.0068
3.0234
3.0244
3.0244
2.5773
2.5772
2.5773
2.5787
2.7213
2.9343
3.0321
3.0493
3.0503
3.0503
2.9671
2.9669
2.967
2.9737
3.0343
3.2329
3.5408
3.5625
3.5637
3.5637
3.0245
3.0244
3.0245
3.0317
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
1.6667
1.6667
1.6667
1.6667
1.6667
1.6667
1.6667
1.6667
1.6667
1.6667
1.8333
1.8333
1.8333
1.8333
1.8333
1.8333
1.8333
1.8333
1.8333
1.8333
2
2
2
2
2
2
2
2
2
2
2.1667
2.1667
2.1667
2.1667
2.1667
2.1667
2.1667
2.1667
2.1667
2.1667
2.3333
2.3333
2.3333
2.3333
0
0.0889
0.1778
0.2667
0.3556
0.4444
0.5333
0.6222
0.7111
0.8
0
0.0889
0.1778
0.2667
0.3556
0.4444
0.5333
0.6222
0.7111
0.8
0
0.0889
0.1778
0.2667
0.3556
0.4444
0.5333
0.6222
0.7111
0.8
0
0.0889
0.1778
0.2667
0.3556
0.4444
0.5333
0.6222
0.7111
0.8
0
0.0889
0.1778
0.2667
2.4176
2.4176
2.4209
2.4398
2.4983
2.6326
2.8107
2.9211
2.9573
2.9635
2.5
2.5
2.5033
2.5224
2.5808
2.7087
2.878
2.9834
3.0181
3.024
2.5
2.5
2.5033
2.5224
2.5807
2.7086
2.8779
2.9833
3.018
3.0239
2.5
2.5
2.5033
2.5224
2.5808
2.7087
2.8781
2.9835
3.0181
3.0241
2.5773
2.5772
2.5772
2.5774
199
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
2.7778
2.7778
2.7778
2.7778
2.7778
2.7778
3
3
3
3
3
3
3
3
3
3
0.4444
0.5556
0.6667
0.7778
0.8889
1
0
0.1111
0.2222
0.3333
0.4444
0.5556
0.6667
0.7778
0.8889
1
3.097
3.3066
3.636
3.9102
4.0774
4.1202
3.0245
3.0244
3.0245
3.0317
3.097
3.3066
3.636
3.9102
4.0774
4.1682
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
2.3333
2.3333
2.3333
2.3333
2.3333
2.3333
2.5
2.5
2.5
2.5
2.5
2.5
2.5
2.5
2.5
2.5
0.3556
0.4444
0.5333
0.6222
0.7111
0.8
0
0.0889
0.1778
0.2667
0.3556
0.4444
0.5333
0.6222
0.7111
0.8
2.5833
2.7213
2.8985
3.0078
3.0437
3.0498
2.8849
2.8848
2.8848
2.8855
2.8951
2.9437
3.0737
3.307
3.3522
3.3599
200
Name
Type
Inputs/Outputs
NumInputMFs
NumOutputMFs
NumRules
AndMethod
OrMethod
ImpMethod
AggMethod
DefuzzMethod
InLabels
OutLabels
InRange
OutRange
InMFLabels
OutMFLabels
InMFTypes
OutMFTypes
24. InMFParams
25. OutMFParams
26.
27.
28.
29.
30.
31.
32.
33.
34.
Rule Antecedent
Rule Consequent
Rule Weight
Rule Connection
System C
mamdani
[2 2]
[3 3]
[2 9]
3
min
max
min
max
centroid
H
ASR
M
n
[1 3]
[1 50]
[1 2]
[2 10]
1 2 3 1 10 50
1 2 3 5 7 4 2 6 8 9 10
trimf
trimf
[0 1 2 0] [1 2 3 0] [2 3 4 0] [-9 1 10 0] [1 10 50
0] [10 50 90 0]
[0 1 2 0] [1 2 3 0] [2 3 4 0] [4 5 6 0] [6 7 8 0]
[3 4 5 0]
[1 2 3 0] [5 6 7 0] [7 8 9 0] [8 9 10 0] [8 9 10
0] [9 10 11 0]
[1 1]
[1 2]
[1 3]
[1 2]
[1 6]
[2 8]
111
111
201
202
203
Name
Type
Inputs/Outputs
NumInputMFs
NumOutputMFs
NumRules
AndMethod
OrMethod
ImpMethod
AggMethod
DefuzzMethod
InLabels
OutLabels
InRange
OutRange
InMFLabels
OutMFLabels
InMFTypes
OutMFTypes
23 InMFParams
24
25
26
27
28
29
30
31
32
33
34
35
36
37
OutMFParams
Rule Antecedent
Rule Consequent
Rule Weight
Rule Connection
ANFIS System D
sugeno
[2 1]
[3 3]
9
9
prod
probor
prod
sum
wtaver
input1
input2
output
[0.5 3.5]
[1 50]
[1.33 1.67]
in1mf1 in1mf2 in1mf3 in2mf1 in2mf2 in2mf3
out1mf1 out1mf2 out1mf3 out1mf4 out1mf5
out1mf6 out1mf7 out1mf8 out1mf9
trimf
linear
[-0.25 0.9883 2.265 0] [1.002 2.236 3.502 0]
[2.211 3.498 4.75 0]
[-23.5 1 25.5 0] [0.9998 25.5 50 0] [25.5 50
74.5 0]
[0.8147 0.04792 0.4771 0]
[0.2715 0.05442 -0.2307 0]
[-0.8405 0.04764 0.1414 0]
[0.898 0.0609 -0.5483 0]
[0.2619 0.06082 -0.6229 0]
[-0.8571 0.06336 0.2343 0]
[0.9153 0.07682 -1.781 0]
[0.2681 0.07687 -1.401 0]
[-0.8753 0.07592 0.7658 0]
[1 1] [1 2] [1 3] [2 1] [2 2] [2 3] [3 1] [3 2] [3 3]
123456789
111111111
111111111
204
1.0000
1.2800
1.5600
1.8300
2.1100
2.3900
2.6700
2.9400
3.2200
3.5000
1.0000
1.2800
1.5600
1.8300
2.1100
2.3900
2.6700
2.9400
3.2200
3.5000
1.0000
1.2800
1.5600
1.8300
2.1100
2.3900
2.6700
2.9400
3.2200
3.5000
1.0000
1.2800
1.5600
1.8300
2.1100
2.3900
2.6700
2.9400
3.2200
3.5000
1.0000
Checking Data
1
1
1
1
1
1
1
1
1
1
6.44
6.44
6.44
6.44
6.44
6.44
6.44
6.44
6.44
6.44
11.89
11.89
11.89
11.89
11.89
11.89
11.89
11.89
11.89
11.89
17.33
17.33
17.33
17.33
17.33
17.33
17.33
17.33
17.33
17.33
22.78
1.33
1.35
1.4
1.46
1.5
1.5
1.5
1.5
1.5
1.5
1.37
1.37
1.4
1.46
1.5
1.5
1.5
1.5
1.5
1.5
1.33
1.35
1.4
1.46
1.5
1.5
1.5
1.5
1.5
1.5
1.36
1.37
1.42
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.41
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
0.5000
0.7200
0.9400
1.1700
1.3900
1.6100
1.8300
2.0600
2.2800
2.5000
0.5000
0.7200
0.9400
1.1700
1.3900
1.6100
1.8300
2.0600
2.2800
2.5000
0.5000
0.7200
0.9400
1.1700
1.3900
1.6100
1.8300
2.0600
2.2800
2.5000
0.5000
0.7200
0.9400
1.1700
1.3900
1.6100
1.8300
2.0600
2.2800
2.5000
0.5000
2
2
2
2
2
2
2
2
2
2
6.78
6.78
6.78
6.78
6.78
6.78
6.78
6.78
6.78
6.78
11.56
11.56
11.56
11.56
11.56
11.56
11.56
11.56
11.56
11.56
16.33
16.33
16.33
16.33
16.33
16.33
16.33
16.33
16.33
16.33
21.11
1.39
1.35
1.33
1.34
1.37
1.41
1.46
1.5
1.5
1.5
1.39
1.36
1.36
1.36
1.37
1.41
1.46
1.5
1.5
1.5
1.39
1.35
1.33
1.34
1.37
1.41
1.46
1.5
1.5
1.5
1.41
1.37
1.35
1.35
1.38
1.43
1.5
1.5
1.5
1.5
1.44
205
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
1.2800
1.5600
1.8300
2.1100
2.3900
2.6700
2.9400
3.2200
3.5000
1.0000
1.2800
1.5600
1.8300
2.1100
2.3900
2.6700
2.9400
3.2200
3.5000
1.0000
1.2800
1.5600
1.8300
2.1100
2.3900
2.6700
2.9400
3.2200
3.5000
1.0000
1.2800
1.5600
1.8300
2.1100
2.3900
2.6700
2.9400
3.2200
3.5000
1.0000
1.2800
1.5600
1.8300
2.1100
22.78
22.78
22.78
22.78
22.78
22.78
22.78
22.78
22.78
28.22
28.22
28.22
28.22
28.22
28.22
28.22
28.22
28.22
28.22
33.67
33.67
33.67
33.67
33.67
33.67
33.67
33.67
33.67
33.67
39.11
39.11
39.11
39.11
39.11
39.11
39.11
39.11
39.11
39.11
44.56
44.56
44.56
44.56
44.56
1.41
1.46
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.48
1.48
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.55
1.55
1.51
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.61
1.61
1.55
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.65
1.64
1.59
1.51
1.5
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
0.7200
0.9400
1.1700
1.3900
1.6100
1.8300
2.0600
2.2800
2.5000
0.5000
0.7200
0.9400
1.1700
1.3900
1.6100
1.8300
2.0600
2.2800
2.5000
0.5000
0.7200
0.9400
1.1700
1.3900
1.6100
1.8300
2.0600
2.2800
2.5000
0.5000
0.7200
0.9400
1.1700
1.3900
1.6100
1.8300
2.0600
2.2800
2.5000
0.5000
0.7200
0.9400
1.1700
1.3900
21.11
21.11
21.11
21.11
21.11
21.11
21.11
21.11
21.11
25.89
25.89
25.89
25.89
25.89
25.89
25.89
25.89
25.89
25.89
30.67
30.67
30.67
30.67
30.67
30.67
30.67
30.67
30.67
30.67
35.44
35.44
35.44
35.44
35.44
35.44
35.44
35.44
35.44
35.44
40.22
40.22
40.22
40.22
40.22
1.4
1.4
1.4
1.41
1.46
1.5
1.5
1.5
1.5
1.47
1.45
1.45
1.45
1.45
1.5
1.5
1.5
1.5
1.5
1.5
1.51
1.51
1.51
1.51
1.5
1.5
1.5
1.5
1.5
1.54
1.57
1.57
1.57
1.56
1.51
1.5
1.5
1.5
1.5
1.57
1.61
1.62
1.62
1.6
206
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
2.3900
2.6700
2.9400
3.2200
3.5000
1.0000
1.2800
1.5600
1.8300
2.1100
2.3900
2.6700
2.9400
3.2200
3.5000
44.56 1.5
44.56 1.5
44.56 1.5
44.56 1.5
44.56 1.5
50 1.67
50 1.65
50 1.6
50 1.54
50 1.5
50 1.5
50 1.5
50 1.5
50 1.5
50 1.5
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
1.6100
1.8300
2.0600
2.2800
2.5000
0.5000
0.7200
0.9400
1.1700
1.3900
1.6100
1.8300
2.0600
2.2800
2.5000
40.22
40.22
40.22
40.22
40.22
45
45
45
45
45
45
45
45
45
45
1.55
1.5
1.5
1.5
1.5
1.6
1.64
1.65
1.65
1.62
1.58
1.52
1.5
1.5
1.5
207
(3.34)
In above figure, the PAPM pulses first gone through integrate and dump detector,
the pulse with maximum amplitude can be identified, then MAP detector can map
the detected pulse to its symbol sequence position (same process as PPM
demodulation). The resulting pulse then feed to a threshold detector to determine
its amplitude levels (same process as PAM demodulation). Thus the received
PAPM pulse can be correctly detected. In equation (3.34), the first part is the
probability for the primary chips, which chooses the maximum value from
208
incoming pulse sequences. The second part calculates the probability of the
secondary pulses in the pulse sequences. To calculate the detection error of the Mn-PAPM modulation, equation (3.36) can be used. Detailed discussion on how to
obtain equation (3.36) and meaning of its variables can be found in Chapter 3.
1
= 1 =
(3.36)
Where =
1
2
( 4 )
(3.30)
< < +
(3.31)
,
+,
<
=
1 ,
,
>0
=0
(3.32)
The MAP detection was same as the PPM case and combined PAPM pulse
detection can be demonstrated in the following figures.
209
In order to simulate the BER for M-n-PAPM modulation scheme under multipath
ISI and artificial light interference, a Matlab program was written with the
following functions:
1. Main_PAPM_default
2. am_prepwere
3. am_vi
4. betaPortion
5. convolve
6. erfh
7. SimulatePAPM
Similar to PAM and PPM case, function 2,3,4,5,6 were public functions and same
as the other two cases. Function 1 and function 7 were different and specific for
PAPM scheme only. The following listed the variable used and details of function
1 and 7.
210
Variables
1. Data rate
Data rate range from 0Mbps to 140Mbps, step of 10Mbps.
2. Ceiling height
Three cases were considered, with H=1, 2, 3 (m)
3. M, n Value
M=1, n=4
4. SNR
SNR=6 dB
5. Artificial light factor
As artificial light was not considered for this case, thus the power ratio ASR was
not enabled.
Functions details
1. Main_PAPM_default
%Calculate BER for PAPM with ISI, and ambient light.
tic %start timer
Rbindex=1;%data rate Rb index
b=0;%initial BER value
global Rb;
for Rb=1e6:10e6:140e6
%=============================
% Noise Model Selection
%----------------------------global shotNoisePresented
shotNoisePresented=1;
global amInteferenceSummationPoints;
%global ambientLightPresented
%ambientLightPresented=0
amInteferenceSummationPoints=2;
%
If this parameter>1 then ambient noise is taken into account,
211
%
and this parameter is a number of integration points.
%
Set this parameter to 1 to disable ambient light.
%
Integration accracy is proportional to this parameter.
%=============================
%=============================
% Default parameters
%----------------------------global ceilingHeight
global Amax
global L
global SNR
global amSAR
global amInterferencePeriodTi
ceilingHeight=1; %Height of the room.
Amax=1; %Number of non-zero amplitude levels.
L=4; %Maximum number of chips in symbol
SNR=6; %Signal To Noise Ratio, db
amInterferencePeriodTi=25.0e-6; %In seconds.
amSAR=0.02; %Signal to Ambient light Ratio. = amSAR = 1/K where K
%is similar parameter from [Wong et al].
%=============================
%=============================
%Derivative parameters:
global a %ISI length parameter in chips. Parameter of h-function.
global SN %SNR not in dB form:
global T %Chip length, seconds.
global avLength %Average number of chips in symbol.
global aphabetCount %Number of symbols in alphabet
global M %Bits per symbol
global bitsPerChip
global scaled_chip_length %T/a
global tapsNumber %"Memory" of multipath channel.
global beta %Discretized h., Array
global lambda %(min non-zero Intensity)/average Intensity:
%=============================
%=============================
% Prepwere parameters
%---------------------------a=2.0*ceilingHeight/3e8;
avLength=L;
aphabetCount=L*Amax; %
M=log(aphabetCount)/log(2.0);
lambda=2*L/(Amax+1); %
%Part II:
212
bitsPerChip=M/avLength;
T=bitsPerChip/Rb;
scaled_chip_length=T/a;
%------------------------------------------------------%estimation of size of sequence beta:
%- - - - - - - - - - - - - - - - - - - - - - - - - - - accuracyEps=1.0e-3;
hThresholdTs= (1.0/accuracyEps)^(1.0/6.0) - 1;
if hThresholdTs<1.0
hThresholdTs=1.0;
end
hThresholdTs;
%mark temporary variable with "w":
wtapsNumber = hThresholdTs/scaled_chip_length;
tapsNumber = (floor(wtapsNumber)) + 1; %1 is taken for safety.
%Convert SN from dB to numbers:
%SNR
SN=exp(SNR/10.0*log(10.0));
%=============================
beta=[1:tapsNumber];
%Create beta:
for k=1:tapsNumber
beta(k)=betaPortion(k-1, scaled_chip_length);
end
beta %display beta value
q=[1:tapsNumber];
%hold on
%b=0;%Initial BER value
am_prepwere();%prepwere Ambient Noise parameters
b(Rbindex)=simulatePAPM();%BER value
%b=simulatePAPM();
%b(Rbindex)=simulatePAPM();
Rbindex=Rbindex+1;%data rate counter
%plot(Rb,b,'rs-');
%semilogy(SNR,b,'rs-');
%hold on;
end
Rb=1:10:140;
semilogy((Rb/(L/(log2(L*Amax)))),b,'rs-');
hold on;
toc
%Find Data Rate with minimum BER value
213
for i=1:length(b)
if b(i)==min(b)
j=i;
end
%i=i+1;
end
Rmin=(Rb(j)/(L/(log2(L*Amax))));
7. SimulatePAPM
%Finds symbol error probability (symbol error rate) for PAPM
%for given channel over all chip sequences, all noise events, and all ambient light
%events.
%It can be speculated that BER_symb=BER/M M - bits/symmbol
%In this procedure, P is denoted as BAmb.
function retv=simulatePAPM()
global a %impulse response parameter
global shotNoisePresented %decision on noise value
global amSAR %signal to artificial light power ratio
global amInteferenceSummationPoints %resolution for artificial interferences
global amInterferencePeriodTi %time variable applied to the lighting model
global Amax % number of amplitude
global L % slot number
global M %number of bits
global SN %signal to noise ratio
global T % integration time period for artificial light model
global tapsNumber %number of pulses selected
global lambda % ratio of peak and average intensity
global scaled_chip_length %number of chips convolved in the channel
%------------------------------------------------%Prevent errors:
%- - - - - - - - - - - - - - - - - - - - - - - - tapsSimulationLimit=10000;
if tapsSimulationLimit<=tapsNumber
message='Taps Number limit exceeded.'
return;
end
if L<2
214
if overFlowProtector>2.0E9
sprintf('symSlotWeight^preSymbolSlots>2E9, ~ bit limit
reached.');
L
AAmax
preSymbolSlots
return;
end
symbol_events=symbol_events*symSlotWeight; %total number
of symbol_events
end
%symbol_events=(L*AAmax)^sslots
%set limit to 1E6
PPMSymbolSimulationLimit=1e6;
%Protect against long calculations:
if PPMSymbolSimulationLimit<=symbol_events
sprintf('Symbol Slots Limit exceeded.');
symbol_events
PPMSymbolSimulationLimit
return;
end
%- - - - - - - - - - - - - - - - - - %Generate symbols and chip sequences.
%------------------------------------%Shortcuts:
amK=1.0/amSAR; %Parameter K-declwered in [7, Wong, ...]
weight_ISI_NOISE=1.0/symbol_events/(symSlotWeight*M);
%Prepwere constants for numerical integration:
GaussNorm=1.0/sqrt(2.0*pi);
216
for e=0:symbol_events-1
%------------------------------------%Generate symbols and chip sequences.
%- - - - - - - - - - - - - - - - - - %
% Symbol is encoded as couple data
%
(d,A) = (PostionOfPrimaryChip, AmplitudeLevelOfPrimaryChip).
% 0<=d<=L-1, 0<A<=Amax
%
% (d,A) is encoded as number sym:
% sym=d*Amax+(A-1)
%
% Hence, sym has range 0<=sym<symSlotWeight=L*Amax.
% In turn, for sequence of Lps symbols, the full number
% of possible sequences is
%
% symbol_events = Lps ^ symSlotWeight
%
% Notations in program:
% preSymbolSlots=Lps
% primaryChip=d
%
mask=symbol_events;%mask=symbol_events=(L*AAmax)^sslots
for slot=0:sslots-1
%sym=rem(mask,L);
sym=rem(mask,symSlotWeight);
%
sym=
((L*AAmax)^sslots)/(L*AAmax)=(L*AAmax)^(sslots-1);
mask=mask-sym;%
mask=mask/L;%
amplitude=rem(sym,L);
primaryChip=(sym-amplitude)/L;
amplitude=amplitude+1;
for i=0:L-1
b(slot*L+i+1)=0;
if primaryChip==i
b(slot*L+i+1)=amplitude;
end
end
end
%- - - - - - - - - - - - - - - - - - %Generate symbols and chip sequences.
%------------------------------------%Cycle through primary chips:
for i=0:L-1
%Fill primary symbol's chips with zeros:
217
for k=0:L-1
b(sslots*L+k+1)=0;
end
for ia=1:Amax
%Make i-th primary chip non-zero:
b(sslots*L+i+1)=ia;
if normedGplus<-9.0
skipSummation=true;
end
if normedGminus>9.0
skipSummation=true;
end
end
if skipSummation
continue;
end
PSuccess=0.0;
% start integration
NIStep= (normedGplus-normedGminus)/NIPoints;
for iNIy=0:NIPoints-1
%Take values in the middle of intervals: add 0.5 to index:
y=(iNIy+0.5)*NIStep+normedGminus;% start integration from
normedGminus to normedGplus with step = NIStep
yWeight=GaussNorm*exp(-y.*y*0.5).*NIStep;
220
References:
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
221
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
[37]
222
[38]
[39]
[40]
[41]
[42]
[43]
[44]
[45]
[46]
[47]
[48]
[49]
[50]
[51]
[52]
[53]
223
[54]
[55]
[56]
[57]
[58]
[59]
[60]
[61]
[62]
[63]
[64]
[65]
[66]
[67]
[68]
[69]
[70]
[71]
[72]
[73]
[74]
[75]
[76]
[77]
[78]
[79]
[80]
[81]
[82]
[83]
[84]
225
[85]
[86]
[87]
[88]
[89]
[90]
[91]
[92]
[93]
[94]
[95]
[96]
[97]
[98]
226
[99]
[100]
[101]
[102]
[103]
[104]
[105]
[106]
[107]
[108]
[109]
[110]
[111]
[112]
[113]
[114]
[115]
[116]
[117]
[118]
[119]
[120]
228